Managing Injuries of the Neck Trial (MINT): a randomised controlled trial of treatments for whiplash injuries

Article history

The research reported in this issue of the journal was commissioned by the HTA programme as project number 02/35/02. The contractual start date was in November 2004. The draft report began editorial review in October 2009 and was accepted for publication in May 2010. As the funder, by devising a commissioning brief, the HTA programme specified the research question and study design. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Copyright statement

© Queen’s Printer and Controller of HMSO 2012. This work was produced by Lamb et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to NETSCC. This journal is a member of and subscribes to the principles of the Committee on Publication Ethics (COPE) (http://www.publicationethics.org/). This journal may be freely reproduced for the purposes of private research and study and may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NETSCC, Health Technology Assessment, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2012 Queen’s Printer and Controller of HMSO

Background

Whiplash injuries (acceleration–deceleration injuries to the neck) are a common type of injury and the sequelae make up a significant proportion of the workload of emergency departments (EDs) as well as having significant implications for orthopaedic services, physiotherapy and primary care. Most often the injury results from rear-end or side-impact motor vehicle collisions, but it may also occur during other activities such as diving. Although whiplash injuries are usually not serious, they may cause substantial long-term morbidity, and because they are so common they are a major health and economic problem around the world. In the UK, the annual cost to the economy was about £2553M in 1990, representing about 18% of the total costs of all road traffic collisions and 0.4% of the gross domestic product. 1 More recent estimates suggest the costs maybe circa £3.1B per annum. 2 The costs are caused mainly by lost productivity owing to absence from work and considerable health service costs involved in the treatment of patients who develop chronic symptoms.

The incidence of whiplash injuries varies around the world, being largely dependent on traffic volumes and road conditions. Not surprisingly, high rates are found in developed countries with high population density and high car ownership. Different studies have estimated incidences of 380 per 100,0003 or 106 per 100,0004 in Australia, 266 to 387 per 100,000 in the USA,5 70 per 100,000 in Quebec,6 and 188 to 325 per 100,000 in the Netherlands. 7 In the UK, there appears to have been a substantial increase in the incidence of whiplash injuries during the 1980s and 1990s and there were around 250,000 new cases in 2003. 1 Similar increases have been reported in Germany8 and the Netherlands. 7 The reasons for the increasing incidence are not completely clear, but there are probably several contributory factors, including increase in traffic volume and, possibly, increased litigation leading to an increase in the reporting of symptoms after road accidents.

Classification of whiplash injuries

Early studies of whiplash used terms such as ‘whiplash’ to mean a variety of different things. Terminology was standardised by the Quebec Task Force (QTF), an international expert group, in 1995. 6 These definitions are now internationally accepted and are used throughout this study. According to the QTF definitions, ‘whiplash’ is the mechanism of injury (acceleration–deceleration injuries usually in the frontal plane), ‘whiplash injuries’ are the soft tissue injuries that result and ‘whiplash-associated disorder (WAD)’ describes the pattern of signs and symptoms that arise. There are five grades of WAD, from grade 0 (no neck complaints or signs) to grade IV (fracture or dislocation) see Table 1. 3

The term ‘late whiplash syndrome (LWS)’ is used throughout the study to describe the chronic complications of whiplash, and defined as: ‘Presence of pain, restriction of motion or other symptoms at 6 months or more after the injury, sufficient to hinder return to normal activities such as driving, usual occupation and leisure’.

Prognostic factors for late whiplash syndrome

The most common symptoms after acute whiplash injury are neck pain and headache, with other symptoms such as neck stiffness, shoulder pain, arm pain or numbness (or both), paraesthesia, weakness, dysphagia, visual and auditory symptoms, dizziness and concentration difficulties occurring in smaller numbers of patients. In most cases, symptoms are short lived, but a substantial minority go on to develop LWS, i.e. persistence of significant symptoms beyond 6 months after injury. Different studies have estimated variable proportions of patients to have persistent symptoms, and the incidence of LWS is not known precisely. In part the differences in reported incidence between studies are due to differences in the definitions of chronic symptoms and differences in the populations studied. However, it is generally accepted that between 20% and 50% of patients will report symptoms persisting for more than 6 months. 9 These patients generate the majority of costs associated with WAD in terms of lost productivity and costs to the health service of providing treatment. 1 Therefore, an important aim in the management of acute WAD injuries is to prevent the development of LWS.

A wide range of factors have been reported to be associated with the development of LWS, including physical factors such as pain intensity and psychological factors such as previous psychological problems, stress and acute psychological response to injury. The existing literature is generally of poor methodological quality and does not provide strong evidence for the association of any risk factor with LWS, but initial pain intensity and neck pain-related disability appear to have the most consistent associations. 10–12 The literature is not of sufficient quality to allow any quantitative assessment of the relative importance of the different risk factors. There is limited evidence for the importance of other risk factors such as post-traumatic stress, self-efficacy and previous chronic pain in development of LWS. 13 Despite considerable research effort, the factors associated with LWS and their relative importance are not well known. Several risk factors identified by various studies are potentially modifiable in the early stages of recovery after whiplash injury, so effective treatment may be able to reduce the proportion of patients who subsequently develop LWS.

Management of whiplash-associated disorders

There are few good-quality randomised trials upon which to base recommendations for practice and the optimum treatment for acute whiplash injury is unknown. In the mid-1990s the QTF review found insufficient evidence supporting the treatments in use at that time. It concluded that promoting activity in the early stages of recovery was probably the most effective strategy, soft collars were not helpful, and physiotherapy, a very common treatment, required rigorous evaluation because of its high costs. The Cochrane review Conservative Treatments for Whiplash14 included 17 studies of treatments for acute whiplash. These evaluated a wide range of treatments, but most were of low methodological quality and differences between them precluded any meta-analyses. Although some individual trials appeared to show superiority of one treatment, results were inconsistent and no clear conclusions could be drawn about the most effective therapy. The authors concluded that large, high-quality trials are needed.

The QTF proposed a ‘stepped care’ clinical pathway in which patients are given advice and education at the initial contact, and then reviewed at 3 weeks. Patients with persisting symptoms would then be provided with more intensive treatment aimed at amelioration of modifiable risk factors for LWS to prevent development of chronic symptoms. Such a strategy is likely to yield the most cost-effective method of providing care for acute whiplash injuries, as it should target intensive (and expensive) treatment to those patients who are most likely to benefit. However, the QTF did not make any recommendations about the specific treatments that should be used.

Another review15 suggested that psychological risk factors predominate as risk factors for poor outcome after whiplash injury, and argue that advice to resume normal activity, using a cognitive behavioural approach, should be the treatment of choice for early management of whiplash. The review was used to develop The Whiplash Book,16 which is a booklet that uses a psychoeducational approach to deliver positive messages about prognosis, pain, returning to normal activities, exercise and self-management of symptoms. The Whiplash Book was developed by an internationally recognised multidisciplinary group. 16 Using recommendations for developing patient information and experience of developing similar material for low back pain,17,18 a draft booklet was peer reviewed by independent multidisciplinary experts. The booklet was formatted by design experts, and tested for proof of concept. 19 The results were promising in so far as patient’s beliefs about pain and poor recovery were modified, and the potential for modifying behaviours on the causal pathway to disability suggested. 19 Disability outcomes were not investigated. Experience gained in back pain research emphasised the importance of practitioner training in using The Whiplash Book to ensure that verbal consultations and written material were consistent. The overall approach is called ‘active management’.

Whether or not The Whiplash Book and active management approach is effective is not yet known. Other systematic reviews have suggested that physical as well as psychological factors may also carry significant amounts of risk for poor outcome,12,20 and some of the evidence used to support the strategy of The Whiplash Book was from the field of low back pain or other chronic conditions, and may not be directly transferable to management of WAD. There are notable differences between WAD and low back pain in terms of the mechanism of injury and psychological consequences. For example, phobic travel anxiety and other psychological manifestations of shock may occur in 50% of people after whiplash injury21,22 but are rarely, if ever, reported in low back pain.

Current UK practice for acute whiplash-associated disorder

The most common treatment for WADs in EDs is advice, but the content and quality of the advice varies. 23 Over 90% of departments suggest using analgesics and gradually increasing movement of the neck. Some departments use soft collars as well, suggesting that they should be removed and the neck exercised on a regular basis. Referral to radiological investigations and physiotherapy occurs in 50% of departments for those patients with more severe symptoms. Physiotherapy is a common treatment for WAD but there are no published data on the types of physiotherapy currently delivered in the UK. The most recent UK guidelines for the physiotherapy management of WAD advocates the use of joint manipulation and mobilisation, soft tissue mobilisations, exercises, education and advice, transcutaneous electrical nerve stimulation (TENS) and multimodal packages. 24

Rationale for Managing Injuries of the Neck Trial

Current treatments for acute WAD are variable and are not supported by good-quality evidence. A stepped care approach to treating whiplash injuries, as proposed by the QTF, is potentially the most cost-effective solution but it requires evaluation and identification of the optimal components of treatment. In this approach, EDs would provide advice on management of WAD. Currently, advice given by EDs in the UK is variable and often contains outdated recommendations. The advice contained in The Whiplash Book may potentially be superior to current standard care, and could become a standard treatment across the NHS. Evaluation is needed to determine whether or not it is beneficial.

The second component of the stepped care pathway is physiotherapy treatment, for patients with persistent symptoms. In the Managing Injuries of the Neck Trial (MINT) we evaluate a package of physiotherapy treatments that are, as far as possible, evidence based and acceptable to physiotherapists and could be implemented throughout the NHS if shown to be effective.

Research objectives

1. To estimate the clinical effectiveness of a stepped care approach over a 12-month period after an acute whiplash injury.

   1. Step 1: The Whiplash Book versus usual care advice (UCA) in EDs.

   2. Step 2: In patients with symptoms persisting at 3 weeks (WAD grades I–III), supplementary treatment comprising either a package of physiotherapy treatments and reinforcement of advice versus reinforcement by a physiotherapist of advice provided at the initial ED contact.

      And: the combined effect of the treatments.

2. To estimate the clinical effectiveness in pre-specified subgroups of patients: those with prior neck problems, psychological or physical risk factors for poor outcome, and those seeking compensation.

3. To estimate the costs of each strategy including treatment and subsequent health-care costs over a period of 12 months and to estimate cost-effectiveness.

4. To gain a participant’s perspective on experiencing a whiplash injury, NHS treatment and recovery within the context of MINT.

Introduction

MINT used a design consisting of two linked randomised controlled trials (RCTs). The first step was a cluster randomised trial in which NHS trusts were randomised to one or other of the ED advice interventions to be compared (The Whiplash Book or UCA), for all patients presenting with acute problems following a whiplash injury. The second step was an individually randomised trial, for patients still experiencing symptoms 3 weeks after their injury, comparing physiotherapy with a single advice session reinforcing the advice given in the ED. The two parts of the trial used a common system of follow-up at 4, 8 and 12 months.

The trial was run in hospitals of 12 NHS acute trusts in the UK: Heart of England NHS Foundation Trust (Heartlands and Solihull Hospitals), North Bristol NHS Trust (Frenchay Hospital), Oxford Radcliffe Hospitals NHS Trust (John Radcliffe Hospital), University Hospitals Coventry and Warwickshire NHS Trust [University Hospital (Walsgrave Site) and Hospital of St Cross, Rugby], Gloucestershire Hospitals NHS Trust (Cheltenham General and Gloucester Royal Hospitals), South Warwickshire General Hospitals NHS Trust (Warwick Hospital), Worcestershire Acute Hospitals NHS Trust (Alexandra Hospital, Redditch and Princess of Wales Community Hospital), University Hospitals Birmingham NHS Foundation Trust (Selly Oak Hospital), Kettering General Hospital NHS Foundation Trust (Kettering General Hospital), Buckinghamshire Hospitals NHS Trust (Stoke Mandeville Hospital), Countess of Chester Hospital NHS Foundation Trust (Countess of Chester Hospital), and Gwent Healthcare NHS Trust (Royal Gwent Hospital, Newport). Some trusts comprised several hospitals and hence some clusters contained more than one ED.

Ethics committee approval

MINT was approved by the Trent Multicentre Research Ethics Committee (MREC, reference MREC/04/4/003), the Local Research Ethics Committee and the Research and Development Committee of each participating centre.

Step 1: Cluster randomised trial of The Whiplash Book versus usual advice

Step 2: Individually randomised trial of physiotherapy versus reinforcement of advice given in emergency departments

Outcome measures and data collection

A common follow-up procedure was used for all participants. Follow-up data collection was by postal questionnaires, completed at 4, 8 and 12 months after ED attendance. The primary outcome was the Neck Disability Index (NDI). The NDI is a self-completed questionnaire that has been used successfully in a postal format in trials of neck treatments. 26,27 It assesses pain-related activity restrictions in 10 areas including personal care, lifting, sleeping, driving, concentration, reading and work, each of which is scored from 0 to 5 (0 = no disability, 5 = total disability).

The SF-12 and EQ-5D were included to assess generic health-related quality of life (HRQoL), and to enable a single utility score for economic evaluation to be derived from the EQ-5D. We used the acute (1 week) recall version of the SF-12 in order to obtain more accurate change in health status. Participants also rated whether or not they had improved, remained the same, or worsened. Resource use was assessed by a short questionnaire that asked about additional NHS or private hospital treatment for the whiplash injury, any GP consultations, manipulation, massage or other treatment. Participants were asked to distinguish between prescription and out-of-pocket expenses.

Follow-up questionnaires were mailed to participants 4, 8 and 12 months after their ED attendance. If questionnaires were not returned within a week, a standardised series of reminders were used:

• questionnaire sent to participant

• questionnaire not returned after 1 week: first telephone prompt

• second copy of questionnaire dispatched if necessary

• questionnaire not returned after a further week: second telephone prompt

• questionnaire not returned: contact participant to collect core outcome data by telephone

• unable to contact participant by telephone: class as ‘non-responder’ and close case.

Follow-ups were classed as ‘closed cases’ when a questionnaire was received from the patient, data were collected by telephone, or the procedure was followed to the end and the patient was classed as a ‘non-responder’.

Participants were asked only at the 12-month follow-up whether or not they had pursued and settled a compensation claim related to their whiplash injury. This was not asked at 4- or 8-month follow-up to avoid stimulation of claims among the trial population.

A research assistant not involved in the recruitment or randomisation processes was responsible for mailing follow-up questionnaires, and for entering responses onto the study database. Blinding of the study team was maintained until final analysis of the data was completed.

Database and data processing

The database was designed and developed in house (Wolfson Institute) using Microsoft Access 2002 (Microsoft Corporation, Redmond, WA, USA). All data were kept and backed up using a LaCie Biometric Safe Drive (London, UK) that was only accessible by authorised staff. For an added security to personal data, the database was encrypted using the encryption software Cryptainer, and Groove Virtual Office software was used to transfer files between secured PCs in the study co-ordinating centre (Warwick Clinical Trials Unit) and the statistical team (Wolfson Institute).

The database imposed rules for data entry which include valid range for responses, linked dates and patient identity among data tables, auto-generated patient identity with check digits, auto-generated dates for despatch records and pop-up warnings for ambiguous entries, which eliminated implausible errors on data entry. Queries were set up to automate checks between linked tables.

Data were single entered into the database by study personnel. The data sets were automatically cleaned on a weekly basis using a computer program to standardise missing and ambiguous responses on follow-up questionnaires according to coding rules developed during the study.

For data quality assessment, 10% of baseline, 5% of 4-month follow-up, 5% of 8-month follow-up, and 10% of 12-month follow-up questionnaires were randomly selected for double entry. We found that the clinical sections of the questionnaires have very low item-level percentage of error and very low overall error rates throughout, less than 5%. The health economics sections suffered some higher percentages of item-level discordance, but most of the discordances were attributed to coding of missing values. Small denominators were also responsible for the higher percentages of item-level discordance. All disagreements found between the two databases during the check were corrected and we also corrected any systematic faults that were detected. We concluded that data quality was good and full double entry was not necessary. In addition, we checked manually for data inconsistencies every 3–4 months using the queries mentioned above, and compared any anomalies found with the paper questionnaires. Amendments were made to the data (both paper and electronic) if necessary.

Statistical analysis

Monitoring

Introduction

The commissioning brief requested a trial of non-surgical, non-pharmacological treatments for acute WAD, applied within the first 6 weeks of injury.

Treatments provided in the early phase of recovery after whiplash injury are targeted to the rapid alleviation of acute symptoms and prevention of LWS. The ranges of treatment possibilities include watchful waiting, advice to promote return of normal activity, psychosocial educational materials, physiotherapy, complementary therapies (not widely available in the NHS), or psychological counselling. More intensive multimodal rehabilitation interventions are applied only when chronic symptoms are established.

The QTF provided useful guidance for determining the treatments to be compared in the trial,6 and suggested that promoting activity in the early stages of the injury was probably the most effective way forward, and that soft collars are not helpful.

The QTF proposed a stepped care clinical pathway in which patients are given advice and education at the initial contact, and then reviewed at 3 weeks and considered for further treatment. The QTF suggest that patients with persisting symptoms should be provided with more intensive treatments by health professionals experienced in the management of WAD. This was the broad schema that we adopted for testing within the trial. At the outset of the trial, we believed that the stepped care model was most likely to yield a cost-effective strategy for management of WAD. However, the QTF did not make precise recommendation about which treatments should be used for people who have persisting symptoms. Physiotherapy is the most widely available treatment for WAD in the UK NHS, and the QTF highlighted the lack of evidence and need for rigorous evaluation of physiotherapy because of the potential high cost. The Cochrane review Conservative Treatments for Whiplash drew similar conclusions. 38

Therefore, we decided to draw the following comparisons:

1. For all participants, either UCA versus an active management advice strategy supplemented by The Whiplash Book at ED attendance.

2. For those participants who had persisting symptoms, a single advice session of physiotherapy versus an intensive physiotherapy programme consistent with dose and method of delivery for the UK NHS.

Advice and usual care interventions tested in Step 1

Advice is considered to be the cornerstone of clinical management for acute WAD. In 1995 the QTF proposed that initial treatment should be advice and reassurance about the favourable prognosis following a whiplash injury. Suggested key messages were that pain is usually short-lived and is controllable, that early return to usual activities with the help of exercises produces a favourable outcome and that the use of soft collars may prolong recovery times.

Subsequent research has suggested that psychological risk factors are important as risk factors for poor outcome, and psychoeducational advice materials [based on cognitive behavioural (CB) model] may prove efficacious in the management of acute WAD outcome. 13,39–41 Prior to and independently of the MINT trial, an expert group developed a psychoeducational booklet for use in acute WAD which has been advocated widely in the UK16 and efficacy trials provided proof of concept evidence that the book challenged and modified people’s thinking about the injury. Experienced gained in back pain research emphasised the importance of practitioner training in using The Whiplash Book to ensure the verbal consultation and written material were consistent. The overall approach is called ‘active management’.

Hence we selected to make two treatment comparisons at the point participants presented to the ED:

1. usual care advice

2. an active management strategy which emphasised the importance of resuming normal activity, the short-lived nature of pain, and positive prognosis after a whiplash injury, using both the consultation with the participants to deliver this message and The Whiplash Book as reinforcement.

This presented a number of challenges. Firstly to determine what usual care is, and secondly, to develop a training package using the active management strategy and The Whiplash Book which promoted routine use within EDs.

Defining usual care

Before starting the trial we undertook a national survey of ED consultants across the UK to estimate the usual care in UK EDs, and a content analysis of advice sheets used. 23

A postal questionnaire was sent to 316 lead consultants from all UK EDs with annual new attendances of > 50,000 people. Consultant leads were asked to indicate the use of a range of treatments and the frequency with which these treatments were used. Samples of written advice were requested and content analysis was conducted and compared with survey responses.

The response rate was 79% (251 of 316). The intervention most frequently used was verbal advice to exercise, reported by 84% of respondents for most or all cases, and advice against the use of a collar (83%). Other treatments reported as being used frequently were written advice and anti-inflammatory medication. One hundred and six consultants (42%) provided a sample of written materials. Reference to expected recovery and encouragement for early return to activities (including work) were included in less than 6% of the written advice leaflets. There were important differences between reported verbal behaviours and written advice. Nearly 50% of written materials suggested the use of a soft collar and contained information on how to use a soft collar, and 61% contained information on solicitors and pursuing a personal injury claim.

Radiological investigations, physiotherapy and analgesics other than non-steroidal anti-inflammatory drugs (NSAIDs) were used only in selected cases, and in only 50% of departments. Case selection criteria were greater pain and more severe symptoms at initial presentation, and those who did not improve.

Of the departments providing written advice to patients, the most common format was material developed by the department [supplied by 176 of 251 (70%) departments to most or all cases]. Only 10 out of the 251 departments (4%) used The Whiplash Book for most or all cases, and the remainder used other formats. The majority of departments who gave advice to exercise did not provide information on exercise in their written information.

Content analysis of the UCA leaflets, demonstrated they did not include information about longer-term recovery, the benefits of early return to activity or the appropriate use of collars.

We concluded that verbal advice is the primary method for managing WAD in EDs and is usually supplemented by written advice. Within individual hospitals there is a lack of consistency between verbal and written advice.

Logan and Holt42 also reported a survey of ED practice for the management of WAD in Wales. They found that most departments (19 of 20) used patient advice sheets, but there was a large disparity in the verbal advice given both within and between departments.

Active management strategy for acute whiplash-associated disorder

We designed a programme to train ED clinicians (medical and other) in the active management strategy. We utilised similar approaches to those found successful in implementing The Back Book and active management strategy for low back pain in primary care. 43,44 We implemented a systematic approach to assessment, and provided regular training to ensure delivery of the active management strategy.

ED staff were asked to emphasise key messages during the consultation as well as give a copy of The Whiplash Book. The messages were:

1. reassurance that prognosis following a whiplash injury is good

2. encouragement to return to normal activities as soon as possible using exercises to facilitate recovery

3. reassurance that pain is normal following a whiplash injury and patients should use analgesia consistently to control this

4. advice against using a collar.

The training programme used behavioural learning theories. 45 We used existing training slots (ED rotational induction or in-service training programmes), using the usual session lengths for each hospital. At most departments training occurred once every 4 months in tandem with junior medical staff rotations. Nursing and allied health professionals working in ED were trained through existing programmes within the department.

The training consisted of a 30- to 40-minute session. Clear objectives were set for clinicians to understand WAD and the need to provide clear, well-articulated messages of the active management strategy alongside The Whiplash Book. Clinicians were asked to use this approach for all patients with WAD grades I–III. 6 The training was delivered by the research team (physiotherapists). If ED staff were unable to take part in training programmes, the research team spoke with clinicians individually and provided summary training sheets.

Training sessions for usual care

We developed a training session for the usual care departments that comprised a session of similar length to the active management session. The content was an educational package about WAD, but included no instructions on the management of WAD. Emphasis was placed on continuing with usual advice for all patients with WAD grades I–III6 and providing participants with information on the trial. The package was delivered in the same way in both arms of the trial.

Comparison of active management and usual care interventions

Advice leaflets from all of the departments who agreed to be randomised into the trial were collected and reviewed prior to randomisation. In comparison to the national survey of ED practice, the leaflets were consistent with the most frequently occurring pattern of advice in the UK (Table 2). The advice sheets were brief; with all being limited to one sheet of A4 or A5 sized paper.

Content analysis of The Whiplash Book demonstrated it to be substantially different to the materials used for usual care, not only in length and detail, but also:

1. – in delivering positive messages about prognosis, promoting the message that pain is nothing to worry about

2. – in promoting early return to normal activities and work. Making recommendations about physical activity, exercise and self-management of symptoms.

A notable exclusion from The Whiplash Book compared with the ED advice leaflets was information on pursuing claims or sponsorship from solicitors.

The Whiplash Book consists of 26 pages of A5 and contains approximately 5000 words and illustrations on every page. 16

The trial team were satisfied that all departments potentially randomised to the usual care arm of the trial would provide what constituted UCA in the UK if they continued to issue patients with the advice leaflet that was normally in use at their department (see Table 2). Departments were monitored for any changes in materials throughout the trial. No changes were detected.

Pilot study (Step 1)

We piloted the active management strategy at one ED from April to August 2005. A video recording of the pilot training session was reviewed by a medical educational specialist and feedback supplied. Communication, structure, content and interaction were evaluated. The overall impression was of a ‘very competent’ teaching session.

Feedback on the training was also sought from the clinicians (mixture of grades) immediately after the training session. All clinicians who were surveyed rated the delivery, content and materials as either good or very good. Minor changes were made to presentation slides and handouts following these evaluations.

Details of the final ED intervention and associated materials are in Appendix 8.

Physiotherapy interventions tested in Step 2

Development of the physiotherapy intervention

We used a number of principles to develop the physiotherapy package:

1. To design an intervention that was reflective of best practice in the UK NHS and was consistent with high-quality, evidence-based clinical guidelines.

2. To ensure the evidence base informed the intervention and discouraged the use of treatments for which there is evidence of no effect. Evidence considered comprised both RCTs, and observational studies of risk factors for poor recovery.

3. To ensure the intervention could be delivered within the context of the UK NHS in terms of staffing and time, and respected physiotherapist’s autonomy in clinical decision-making.

4. To ensure the intervention was documented to a standard that promoted consistency in delivery, and would enable replication.

This was achieved by a triangulation of methods (Figure 1) – including systematic reviews of the RCTs and observational studies, a review of clinical guidelines, expert peer review and piloting to test the feasibility and acceptability of the intervention.

FIGURE 1.

Intervention design considerations.

Comparator treatment: reinforcement of advice

The control intervention was a single advice session that aimed to re-enforce the advice provided in the ED. The advice was given by a physiotherapist, and included a brief assessment of symptoms and movements to allow individual tailoring of advice. For example, the physiotherapist could highlight a particular exercise on the advice sheet and advise the patient about performing it, but they could not prescribe a new exercise.

Pilot study

The acceptability and feasibility of the intervention was tested in a pilot study that ran from May to September 2005 and involved three senior outpatient physiotherapists. The physiotherapists attended a half-day training session and each received a comprehensive manual describing the intervention. The physiotherapy package was then delivered to seven patients who fulfilled the criteria for the trial. The treating physiotherapists provided feedback on the training and the intervention. The main feedback from the physiotherapists was that they lacked confidence in the use of the psychological strategies. More time was allocated to this in subsequent training sessions.

Peer review

Following the pilot study minor modifications were made to the intervention manual and reviewed by two experienced researchers (Christopher McCarthy and Michele Sterling, one of whom was involved in research into WAD) and an experienced senior physiotherapist (Heidi Williams). Minor changes were made in response to their feedback.

Physiotherapist training

All physiotherapists received training in the package. Training was held over 1 or 1.5 days depending on the number of physiotherapists attending. The research team physiotherapists provided the training. Participating physiotherapists were provided with a training manual that contained details about the trial, physiotherapy assessment, treatment planning and treatments.

We also provided training on the advice session, including role-play. Guidelines for the delivery of the advice session were provided in the training manual.

Physiotherapist support

The MINT team were in contact with the treating physiotherapists throughout the duration of the trial. An update evening was held in November 2006 to provide a forum for physiotherapists to discuss any problems that had arisen. These were held at 2 centres (Frenchay Hospital and the University of Warwick) and were attended by 23 physiotherapists from 9 out of 11 trusts that were recruiting (Stoke Mandeville had yet to start recruitment). The physiotherapists from the two trusts that were unable to attend were visited to provide an update on the trial.

Introduction

The results section is structured to present the results of the cluster randomised trial of ED advice interventions (Step 1) first, followed by the results of the individually randomised comparison of physiotherapy versus no physiotherapy (Step 2).

Step 1

Recruitment to Step 1

During the recruitment period ED staff completed trial proformas for 7702 patients; 3034 from UCA departments and 4668 from WBA departments (Figure 2).

FIGURE 2.

CONSORT flow diagram for Step 1.

The results of audits conducted to monitor the proportion of patients referred to the trial (see Table 6) are summarised in Table 5. It was not possible to conduct audits in three departments. The average period of audit over the course of recruitment was 10 weeks per hospital. Just under half of all eligible patients attending a participating ED were reported to the trial through completion of a trial proforma. The proportion was slightly higher in the UCA centres compared with the WBA centres, but this was offset by the slightly higher numbers of eligible patients seen per day in the WBA centres (2.7 vs 2.3). There were no major differences between the trial arms in the proportion of eligible patients referred to the trial. Audit of the number of proformas completed and the number of advice leaflet packs given out showed that in all centres (where audits were possible), except one, the number of information packs given out exceeded the number of patients referred to the trial (Table 6). This suggests that virtually all patients for whom a proforma was completed also received the information pack, but also that some patients who were not referred to the trial also received an information pack (340 proformas were completed and 372 advice leaflet packs were used).

TABLE 5 - Attendance rates and proforma completion rates at UCA and WBA centresa

Data unavailable for one UCA and one WBA centre.

These centres were trusts with two EDs.

Centre	Average daily attendances (eligible patients per day)	% patients with proforma completed
WBA centres
University Hospitals Birmingham NHS Foundation Trustb	2.9	38
University Hospitals Birmingham NHS Foundation Trustb	2.2	36
Oxford Radcliffe Hospitals NHS Trust	2	49
Gwent Healthcare NHS Trust	1.8	48
University Hospitals Birmingham NHS Foundation Trust	2.8	54
South Warwickshire General Hospitals NHS Trust	1.8	51
Average daily attendance per cluster	2.7	46
UCA centres
University Hospitals Coventry and Warwickshire NHS Trustb	3.2	37
University Hospitals Coventry and Warwickshire NHS Trustb	1.3	54
Gloucestershire Hospitals NHS Trustb	1.5	56
Gloucestershire Hospitals NHS Trustb	0.6	61
Kettering General Hospital NHS Foundation Trust	1.5	46
North Bristol NHS Trust	1.9	28
Worcestershire Acute Hospitals NHS Trust	1.7	61
Average daily attendance per cluster	2.3	49

TABLE 6 - Results of audits of information packs

These centres were trusts with two EDs.

Centre	No. of days audited	No. of proformas received	No. of information packs given out
WBA centres
University Hospitals Birmingham NHS Foundation Trusta	24	42	44
University Hospitals Birmingham NHS Foundation Trusta	12	15	21
Oxford Radcliffe Hospitals NHS Trust	70	46	58
Gwent Healthcare NHS Trust	14	11	11
University Hospitals Birmingham NHS Foundation Trust	13	31	31
South Warwickshire General Hospitals NHS Trust	40	31	35
UCA centres
North Bristol NHS Trust	161	84	92
Worcestershire Acute Hospitals NHS Trust	13	11	12
University Hospitals Coventry and Warwickshire NHS Trusta	36	60	58
University Hospitals Coventry and Warwickshire NHS Trusta	32	9	10

Of the 7702 patients referred to the trial, 6952 were eligible and were approached to be included in the trial by sending the trial 2-week questionnaire. A total of 3851 (55%) patients agreed to participate by returning the questionnaire: 2253 from WBA centres and 1598 from UCA centres. The number of patients recruited by each centre is given in Table 7. More patients were recruited to centres in the WBA arm than to centres in the UCA arm (Figure 3), which was a result of a higher recruitment rate (per 10,000 attendances) in the WBA arm (Figure 4).

TABLE 7 - Characteristics of clusters

Cluster number	Trust	Start date	Months of recruitment	Size (attendances/year)	Numbers recruited (recruits per 10,000 attendances)
UCA
1	University Hospitals Coventry and Warwickshire NHS Trust	17 January 2006	22	145,801	370 (25)
2	North Bristol NHS Trust	9 December 2005	23	89,829	228 (25)
3	Gloucestershire Hospitals NHS Trust	9 February 2006	21	127,461	431 (34)
4	Worcestershire Acute Hospitals NHS Trust	11 April 2006	19	49,524	302 (61)
5	Kettering General Hospital NHS Foundation Trust	22 May 2006	17	65,022	212 (33)
6	Buckinghamshire Hospitals NHS Trust	6 December 2006	12	27,929	55 (20)
WBA
7	Heart of England NHS Foundation Trust	6 December 2005	23	140,388	711 (51)
8	University Hospitals Birmingham NHS Foundation Trust	5 December 2005	23	76,845	524 (68)
9	Oxford Radcliffe Hospitals NHS Trust	8 February 2006	21	117,073	224 (19)
10	South Warwickshire General Hospitals NHS Trust	24 April 2006	19	53,008	252 (48)
11	Gwent Healthcare NHS Trust	21 September 2006	13	70,554	130 (18)
12	Countess of Chester Hospital NHS Foundation Trust	31 July 2006	15	56,240	412 (73)

FIGURE 3.

Observed versus expected cumulative patient recruitment by trial arm. The expected number of recruits was reset to the actual observed number at the time of the extension to recruitment, hence the discontinuity in the ‘expected’ line.

FIGURE 4.

Average monthly recruitment rates by trial arm.

The characteristics of patients recruited to the trial are summarised in Table 8. The WBA and UCA arms were well matched but there were small differences between them in ethnicity; there were slightly more people of Pakistani origin in the WBA group than in the UCA group (8% vs 2%), and fewer people of white ethnic origin (70% vs 84%). This difference reflects differences in the populations served by the hospitals in the two arms. The majority of the participants were of working age (mean age 37 years, SD 13), and were employed. There was a small difference between the arms in the proportion that were working (UCA 75%, WBA 69%). The age distributions of the participants were similar in the two arms (Figure 5), with a peak between 20 and 30 years. Road traffic accidents were by far the most common mechanism of injury, with other causes accounting for only 5% of the participants (200 of 3851). WAD grade III injuries (involving neurological signs as well as pain and musculoskeletal signs) were rare (104 of 3851, 2.7% of participants).

TABLE 8 - Characteristics of participants by trial arm. Mechanism of injury, location of pain, pain intensity, medical history, and WAD grades were collected at EDs attendance. Gender, age, and ethnic group were collected on the 2-week follow-up questionnaire

Full-time employed, part-time employed, and self-employed.

Unemployed, full-time student, and retired/looking after home/inactive.

	UCA	Missing	WBA	Missing
Number enrolled	1598		2253
Gender – male	666 (42%)	18	995 (44%)	39
Age in years, mean [SD]	37 [13]	0	37 [13]	0
Ethnic group		118		224
White	1336 (84%)		1586 (70%)
Mixed	19 (1%)		42 (2%)
Indian	49 (3%)		95 (4%)
Pakistani	24 (2%)		179 (8%)
Bangladeshi	9 (1%)		21 (1%)
Black or Black British	31 (2%)		69 (3%)
Chinese or other	12 (1%)		37 (2%)
Mechanism of injury		15		14
Road traffic accident	1495 (94%)		2127 (94%)
Other	88 (6%)		112 (5%)
Location of pain		37		73
C-spine only	1046 (65%)		1365 (61%)
C-spine and other spinal area	275 (17%)		400 (18%)
Other spinal area only	31 (2%)		65 (3%)
Spinal and other area	141 (9%)		190 (8%)
Other area only	23 (1%)		56 (2%)
No pain	45 (3%)		104 (5%)
Pain intensity (/10), mean [SD]	4.9 [1.9]	349	5.3 [1.9]	574
History
Previous neck problems	190 (12%)	58	218 (10%)	94
Previous back problems	199 (12%)	308	285 (13%)	396
Neurological symptoms	98 (6%)	57	121 (5%)	86
WAD grades		0		0
I: Complaint of pain, stiffness or tenderness, no physical signs	883 (55%)		1205 (53%)
II: Complaint of pain, stiffness or tenderness, musculoskeletal signs	662 (41%)		997 (44%)
III: Complaint of pain, stiffness or tenderness, neurological signs	53 (3.3%)		51 (2.3%)
Employment		155		272
Working/earninga	1185 (74%)		1549 (69%)
Unpaid work	4 (0.25%)		7 (0.31%)
Not workingb	254 (16%)		425 (19%)

FIGURE 5.

Age at ED attendance.

Follow-up questionnaires

The chronology of trial questionnaires for a participant in Step 1 is summarised in Figure 6.

FIGURE 6.

Step 1: chronology of trial questionnaires.

Because many participants did not return their follow-up questionnaires immediately the actual times until questionnaire return were slightly longer than the specified follow-up periods (Table 9).

TABLE 9 - Median time (in months) from ED attendance to questionnaire completion

IQR, interquartile range.

	UCA Median (IQR)	WBA Median (IQR)
Time to 2-week follow-up	0.5 (0.4–0.8)	0.5 (0.3–0.8)
Time to 4-month follow-up	4.3 (4.1–4.9)	4.3 (4.1–5.0)
Time to 8-month follow-up	8.4 (8.1–9.0)	8.4 (8.1–9.0)
Time to 12-month follow-up	12.4 (12.2–13.1)	12.5 (12.2–13.2)

In total 197 participants withdrew from the trial during follow-up, and were not sent any further follow-up questionnaires after their withdrawal. Withdrawals were well balanced between the trial arms: 39 UCA (2.4%) and 46 WBA (2.0%) participants withdrew after returning the 2-week questionnaire, 37 UCA (2.3%) and 36 WBA (1.6%) participants withdrew after the 4-month questionnaire, and 22 UCA (1.4%) and 17 WBA (0.1%) participants withdrew after the 8-month questionnaire. A reason for withdrawal was given by the majority of participants who withdrew (Table 10). There were no serious adverse events reported in Step 1 of the trial.

TABLE 10 - Reasons for withdrawal by trial arm

Reasons for withdrawal	UCA	WBA	Total
Illness of family member	0	1	1
Inconvenient to attend treatment	0	1	1
Language barrier	0	1	1
Moved away	4	11	15
No longer interested	60	52	112
No reasons given	3	2	5
Patient died – unrelated to trial	0	4	4
Post inclusion ineligibility	2	2	4
Recovered and no longer interested	11	11	22
Self-reported ineligibility	9	6	15
Unhappy with NHS service	1	2	3
Unhappy with trial	8	6	14
Total	98	99	197

The proportion of participants providing outcome data was 80% at 4 months, 71% at 8 months and 70% at 12 months (Figure 2). Non-response was the major cause of missing outcome data. There was no evidence of major differences in age, sex or injury severity between participants retained by the trial at 12 months and those lost to follow-up (Table 11), though there was a slightly higher proportion of male participants among those lost.

TABLE 11 - Characteristics of patients retained at 12 months and patients who were lost to follow-up

	Patients retained at 12 months	Patients lost to follow-up
UCA	n = 1127	n = 471
Age, mean (SD)	39 (14)	34 (13)
Male, n (%)	438 (39)	228 (48)
WAD I, n (%)	631 (56)	252 (54)
WAD II, n (%)	459 (41)	203 (43)
WAD III, n (%)	37 (3)	16 (3)
WBA	n = 1577	n = 676
Age, mean (SD)	38 (14)	34 (13)
Male, n (%)	681 (43)	314 (47)
WAD I, n (%)	862 (55)	343 (51)
WAD II, n (%)	684 (43)	313 (46)
WAD III, n (%)	31 (2)	20 (3)

Comparison of The Whiplash Book advice and usual care advice arms

The tables in this section present the results of the analyses described in Chapter 2, Statistical analysis. Each table includes two analyses: the complete case analysis (i.e. using all participants who provided outcome data) and the multiply imputed analyses, which allows for the estimated effects of missing data. All analyses were adjusted for WAD grade at ED clustering of NHS trusts. Estimates of the ICC for each outcome are presented in Appendix 3. The ICC estimate for the primary outcome (NDI) was 0.0136 at 4 months and substantially lower at 8 and 12 months (0.0075 and 0.0033, respectively).

Primary outcomes: Neck Disability Index

There was no evidence of a difference in NDI score between the two ED advice intervention arms (Table 12 and Figure 7). Because of the imbalance in ethnicity between the advice interventions, we reran the analysis on complete cases with additional adjustment for ethnicity, as a sensitivity analysis. There was no material difference in the treatment effect estimates (Table 13). A further sensitivity analysis restricted to white participants only, produced very similar results.

TABLE 12 - Neck Disability Index score (%)

Complete case estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed estimate adjusted for clustering and WAD grade at ED attendance.

	4 months	8 months	12 months
UCA
Mean (SD)	20.4 (17.2)	16.0 (16.4)	14.4 (16.0)
n (missing/total) (%)	42/1,295 (3%)	36/1175 (3%)	25/1127 (2%)
WBA
Mean (SD)	21.5 (17.6)	16.6 (16.5)	14.4 (15.9)
n (missing/total) (%)	64/1774 (4%)	35/1570 (2%)	41/1577 (3%)
Treatment estimatea
∆ (95% CI)	0.5 (–1.8 to 2.8)	0.3 (–1.6 to 2.2)	–0.1 (–1.6 to 1.4)
Treatment estimateb
∆ (95% CI)	0.5 (–2.1 to 3.0)	0.8 (–1.6 to 3.1)	0.5 (–1.5 to 2.5)

FIGURE 7.

Observed NDI scores (%) (mean and 95% CIs) by advice intervention. Higher score indicates greater disability.

TABLE 13 - Neck Disability Index score (%), further adjusted for ethnicity

Complete cases estimate adjusted for clustering, WAD grade at ED attendance and ethnicity.

	4 months	8 months	12 months
Treatment estimatea
∆ (95% CI)	–0.4 (–1.7 to 0.9)	–0.5 (–2.5 to 1.5)	–0.6 (–2.9 to 1.6)

Secondary outcomes

There were no statistically or clinically significant differences between the UCA and WBA groups for the physical or mental health components of the SF-12 (Tables 14 and 15, Figures 8 and 9). We also reran these analyses by further adjusting the regressions for the effect of ethnicity, because of the observed imbalance. This did not alter the results for either the mental or physical components of the SF-12 (see Appendix 4).

TABLE 14 - Mental component score of SF-12v1

Complete cases estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed estimate adjusted for clustering and WAD grade at ED attendance.

MCS	2 weeks	4 months	8 months	12 months
UCA
Mean (SD)	40.9 (13.0)	48.0 (11.6)	49.4 (11.3)	49.6 (10.9)
n (missing/total) (%)	183/1488	256/1295	266/1175	277/1127
WBA
Mean (SD)	40.7 (12.7)	47.1 (12.1)	48.7 (11.5)	49.3 (10.9)
n (missing/total) (%)	271/2042	406/1774	359/1570	487/1577
Treatment estimatea
∆ (95% CI)	–0.1 (–1.8 to 1.5)	–0.9 (–1.9 to 0.0)	–0.8 (–1.8 to 0.2)	–0.3 (–1.3 to 0.7)
Treatment estimateb
∆ (95% CI)	–0.4 (–1.9 to 1.2)	–0.3 (–1.6 to 1.0)	–0.5 (–1.7 to 0.7)	–0.3 (–1.4 to 0.9)

TABLE 15 - Physical component score of SF-12v1

Complete cases estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed estimate adjusted for clustering and WAD grade at ED attendance.

PCS	2 weeks	4 months	8 months	12 months
UCA
Mean (SD)	40.3 (9.0)	46.5 (9.7)	48.9 (9.2)	49.9 (9.0)
n (missing/total) (%)	183/1488	256/1295	266/1175	277/1127
WBA
Mean (SD)	40.2 (8.9)	46.0 (9.8)	48.5 (9.2)	49.8 (9.1)
n (missing/total) (%)	271/2042	406/1774	359/1570	487/1577
Treatment estimatea
∆ (95% CI)	–0.3 (–2.3 to 1.7)	–0.4 (–1.9 to 1.2)	–0.4 (–1.2 to 0.4)	–0.2 (–1.6 to 1.2)
Treatment estimateb
∆ (95% CI)	–0.3 (–2.2 to 1.5)	–0.5 (–2.0 to 1.1)	–0.0 (–1.1 to 1.0)	0.0 (–1.5 to 1.5)

FIGURE 8.

Observed mental component scores of SF-12v1 (mean and 95% CI) by advice intervention. Higher scores indicate better mental health-related quality of life.

FIGURE 9.

Observed physical component scores of SF-12v1 (mean and 95% CI) by advice intervention. Higher scores indicate better physical health-related quality of life.

There were no statistically significant differences between the UCA and WBA groups for acute whiplash injury and LWS (Tables 16 and 17).

TABLE 16 - Acute whiplash injury

Complete cases odds ratio estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed odds ratio estimate adjusted for clustering and WAD grade at ED attendance.

	n (%) or odds ratio (OR) (95% CI)
UCA
Present	492 (31%)
Absent	796 (50%)
n (missing/total) (%)	310/1598 (19%)
WBA
Present	704 (31%)
Absent	1061 (47%)
n (missing/total) (%)	488/2253 (22%)
Treatment estimatea
∆ (OR 95% CI)	1.00 (0.76 to 1.33)
Treatment estimateb
∆ (OR 95% CI)	1.01 (0.77 to 1.33)

TABLE 17 - Late whiplash syndrome

Complete cases odds ratio estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed odds ratio estimate adjusted for clustering and WAD grade at ED attendance.

	n (%) or odds ratio (OR) (95% CI)
UCA
Present	286 (18%)
Absent	839 (53%)
n (missing/total) (%)	473/1598 (30%)
WBA
Present	397 (18%)
Absent	1176 (52%)
n (missing/total) (%)	680/2253 (30%)
Treatment estimatea
∆ (OR 95% CI)	0.95 (0.73 to 1.25)
Treatment estimateb
∆ (OR 95% CI)	1.05 (0.86 to 1.29)

There were no statistically significant differences between the UCA and WBA groups for work days lost (Tables 18–20).

TABLE 18 - Work days lost (cumulative) in Step 1 (Poisson regression)

Complete cases analysis; percentage difference adjusted for ED clustering, WAD grade at baseline and time from ED attendance.

Multiple imputation analysis; percentage difference adjusted for ED clustering, WAD grade at baseline and time from ED attendance.

	4 months	8 months	12 months
UCA
Mean (SD)	4.4 (13.1)	5.1 (14.5)	5.6 (15.8)
Missing/total	9/1295 (1%)	0/1175 (0%)	2/1127 (0%)
WBA
Mean (SD)	4.4 (13.5)	5.4 (15.9)	5.8 (17.4)
Missing/TOTAL	8/1774 (0%)	4/1570 (0%)	3/1577 (0%)
Treatment effect
Δ (95% CI)a	–7.5 (–26.8 to 17.0)	–24.4 (–62.8 to 53.4)	–21.3 (–42.3 to 7.3)
Δ (95% CI)b	–7.0 (–26.0 to 17.0)	–24.3 (–62.6 to 53.2)	–18.4 (–39.7 to 10.4)

TABLE 19 - Work days lost (cumulative) in Step 1 additionally adjusted for not working at baseline (Poisson regression)

Complete cases analysis; percentage difference adjusted for ED clustering, WAD grade at baseline and time from ED attendance.

Multiple imputation analysis; percentage difference adjusted for ED clustering, WAD grade at baseline and time from ED attendance.

	4 months	6 months	12 months
Δ (95% CI)a	–15.2 (–33.9 to 8.7)	–24.5 (–63.8 to 57.5)	–23.7 (–45.4 to 6.5)
Δ (95% CI)b	–14.4 (–32.8 to 9.2)	–24.1 (–63.5 to 57.9)	–20.3 (–42.4 to 10.2)

TABLE 20 - Work days lost (cumulative) in Step 1: restricted analysis of those who worked at baseline (Poisson regression)

Complete cases analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

Multiple imputation analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

	4 months	8 months	12 months
UCA
Mean (SD)	5.3 (14.3)	6.2 (15.9)	6.7 (17.5)
Missing/total	5/950 (1%)	0/901 (0%)	0/863 (0%)
WBA
Mean (SD)	4.7 (13.1)	5.8 (16.0)	6.3 (17.8)
Missing/total	1/1194 (0%)	1/1131 (0%)	2/1134 (0%)
Treatment effect
Δ (95% CI)a	–17.4 (–35.3 to 5.5)	–26.4 (–64.5 to 52.3)	–26.8 (–47.3 to 1.7)
Δ (95% CI)b	–16.3 (–34.3 to 6.5)	–25.8 (–64.2 to 53.5)	–23.6 (–44.6 to 5.5)

Subgroup analyses

Three subgroup analysis of the primary outcome are presented, stratifying by injury severity (WAD grade at ED presentation), previous neck pain, and early psychological response to whiplash injury. The analysis of injury severity was not adjusted for WAD grade, as this was the factor used to classify the subgroups. Otherwise adjustments were consistent with previous analyses and are detailed in footnotes of each table.

Tables 21–23 provide the estimates for the NDI in each subgroup and the statistical significance of interaction terms at each time point of follow-up. There were no statistically significant differences between the subgroups in the effects of the intervention. Estimates [mean and 95% confidence interval (CI)] of all terms tested in the NDI subgroup analyses are given in Appendix 5.

TABLE 21 - Neck Disability Index and injury severity

Complete cases estimate adjusted for clustering.

Multiply imputed estimate adjusted for clustering.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Severe symptoms
UCA	22.5 (17.9) [540]	18.5 (17.6) [498]	16.1 (16.9) [479]
WBA	23.5 (18.1) [793]	18.1 (17.1) [714]	16.0 (16.8) [695]
Milder symptoms
UCA	18.8 (16.4) [713]	14.0 (15.2) [641]	13.1 (15.1) [623]
WBA	19.8 (17.1) [917]	15.2 (15.9) [821]	13.2 (14.9) [842]
p-value for interactiona	0.96	0.22	0.88
p-value for interactionb	0.94	0.90	0.66

TABLE 22 - Neck Disability Index and early psychological response

Complete cases estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed estimate adjusted for clustering and WAD grade at ED attendance.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Early psychological response present
UCA	25.6 (17.8) [644]	20.8 (18.0) [606]	18.5 (17.6) [563]
WBA	26.4 (17.7) [867]	20.9 (17.4) [808]	18.3 (16.9) [787]
Early psychological response absent
UCA	14.0 (14.2) [492]	10.2 (12.0) [475]	9.7 (12.3) [460]
WBA	14.0 (13.7) [634]	10.5 (12.1) [617]	9.3 (12.1) [622]
p-value for interactiona	0.56	0.91	0.82
p-value for interactionb	0.73	0.79	0.81

TABLE 23 - Neck Disability Index and previous neck pain

Complete cases estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed estimate adjusted for clustering and WAD grade at ED attendance.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Had previous neck pain
UCA	27.1 (18.2) [62]	23.4 (17.7) [61]	21.6 (17.9) [64]
WBA	31.7 (18.1) [93]	25.5 (17.3) [89]	21.6 (18.2) [84]
No previous neck pain
UCA	20.1 (17.0) [1077]	15.7 (16.2) [1022]	14.0 (15.7) [961]
WBA	20.5 (17.1) [1407]	15.7 (15.8) [1336]	13.9 (15.5) [1314]
p-value for interactiona	0.18	0.49	0.93
p-value for interactionb	0.24	0.61	0.87

p-value for interaction is the interaction between WBA and injury severity. No adjustment for WAD grade at ED attendance was made here because of collinearity (injury severity was determined from WAD grade at ED attendance).

Additional subgroup analyses were performed for the SF-12, LWS and acute whiplash indicator. There were no statistically significant differences between the subgroups in the effects of the interventions, for any of these other outcomes. A few of the interaction tests had p-values between 0.05 and 0.10, but given the large number of tests performed these are likely to have been due to chance.

Compensation pursued

We also performed a subgroup analysis stratifying patients by whether or not they were pursuing or had pursued a claim for compensation in relation to their whiplash injury. This information was collected only at the 12-month follow-up to avoid stimulating compensation claims among the trial population. Hence this is not a true subgroup analysis, as it does not classify participants by a baseline characteristic; the compensation claim would certainly have been initiated after injury, and may also have been after enrolment in the trial.

Table 24 presents the characteristics of participants who did and did not pursue a compensation claim (as reported at 12 months). There were few participants who did not pursue a compensation claim in either trial arm, hence there is limited power to detect any differences between those who did and did not pursue a claim. There was no evidence that the ED advice intervention had any effect on the number of participants pursuing a claim (Table 25), nor that its effects differed between those who pursued a compensation claim and those who did not (Table 26).

TABLE 24 - Characteristics of those that claimed compensation

	UCA	Missing, n	WBA	Missing, n
Pursued compensation	1037		1441
Male, n (%)	401 (39%)	12	626 (43%)	16
Age in years, mean (SD)	39 (13)	0	38 (13)	0
Injury severity, n (%)		0		0
WAD I	584 (56%)		787 (55%)
WAD II	418 (40%)		626 (43%)
WAD III	35 (3%)		28 (2%)
Medium used to claim	643 (62%)	394	936 (65%)	505
Through solicitor	424 (41%)		582 (40%)
Directly from insurer	205 (20%)		328 (23%)
Other medium	14 (1%)		26 (2%)
Claim has been resolved	375 (36%)	3	560 (39%)	4
Did not pursue compensation	63		93
Male, n (%)	23 (37%)	2	36 (39%)	2
Age in years, mean (SD)	37 (15)	0	33 (12)	0
Injury severity, n (%)	63 (100%)	0	93 (100%)	0
WAD I	33 (52%)		51 (55%)
WAD II	28 (44%)		40 (43%)
WAD III	2 (3%)		2 (2%)
Missing compensation status		498		719

TABLE 25 - Comparison of compensation between trial arms

Complete cases estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed estimate adjusted for clustering and WAD grade at ED attendance.

	n (%) or odds ratio (95% CI)
UCA
Pursued	1037 (65%)
Did not pursue	63 (4%)
n (missing/total) (%)	498/1598 (31%)
WBA
Pursued	1441 (64%)
Did not pursue	94 (4%)
n (missing/total) (%)	718/2253 (32%)
Treatment estimatea
∆ (95% CI)	1.21 (0.31 to 4.81)
Treatment estimateb
∆ (95% CI)	1.21 (0.31 to 4.81)

TABLE 26 - Neck Disability Index and compensation

Complete cases estimate adjusted for clustering and WAD grade at ED attendance.

Multiply imputed estimate adjusted for clustering and WAD grade at ED attendance.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Pursued compensation
UCA	20.2 (17.1) [925]	16.3 (16.4) [900]	14.9 (16.1) [1017]
WBA	20.5 (17.2) [1249]	16.1 (15.9) [1216]	14.7 (16.1) [1412]
Did not pursue compensation
UCA	11.6 (15.9) [42]	10.2 (13.7) [47]	7.2 (10.3) [62]
WBA	14.4 (15.6) [67]	9.8 (12.9) [71]	8.7 (10.8) [93]
p-value for interactiona	0.56	0.90	0.62
p-value for interactionb	0.83	0.89	0.48

Step 2

Recruitment to Step 2

Recruitment to Step 2 took place concurrently with recruitment to Step 1, from December 2005 to November 2007. Although we expected that all participants in Step 2 would already be taking part in Step 1, in fact 25 patients were randomised into Step 2 before returning their 2-week questionnaire, and subsequently failed to return either this or the 4-month questionnaire. Hence these 25 patients were randomised into Step 2 and are included in the results, but were not included in the earlier follow-up stages of Step 1.

A total of 949 patients reported on-going problems to the trial office, and were considered for the second stage of the trial. Of these, 693 were assessed as potentially eligible and were invited to a research clinic appointment. Seventy-seven did not attend or cancelled, and 616 patients were assessed for eligibility, of whom 2 were ineligible, 15 declined participation and 599 were randomised. No patients were excluded because of contraindications to physiotherapy.

Table 27 summarises the characteristics of the participants randomised to Step 2 and those who were not randomised. The differences were not large, but the population randomised to Step 2 included a slightly lower proportion of males, a higher proportion of people with more serious injuries (WAD grade II or III) at ED attendance, and lower SF-12 scores.

TABLE 27 - Characteristics of patients randomised to Step 2 versus those who were not randomised

Whiplash-associated disorder grades 0 and IV were not eligible for the trial.

	Randomised	Missing	Not randomised	Missing
Number of patients	599	0	3277	0
Sex – male	221 (37%)	0	1456 (44%)	50
Age in years, mean [SD]	40 [13]	0	36 [13]	0
Had previous neck pain	77 (13%)	40	334 (10%)	115
WAD grades		0		0
0: No neck complaints or signsa	0 (0%)		0 (0%)
I: Complaints of pain, stiffness or tenderness, no physical signs	275 (46%)		1823 (56%)
II: Complaint of pain, stiffness or tenderness, musculoskeletal signs	299 (50%)		1375 (42%)
III: Complaint of pain, stiffness or tenderness, neurological signs	25 (4%)		79 (2%)
IV: Fracture/dislocationa	0 (0%)		0 (0%)
SF-12v1 scores, mean [SD]
Mental component score	36 [12]	108	42 [13]	692
Physical component score	36 [7]	108	41 [9]	692
Received public fund	192 (33%)	37	748 (23%)	332

Previous neck pain and WAD grades were collected at ED attendance. Gender, age, MCS, and PCS were collected via the 2-week follow-up questionnaire.

Participants were recruited from the two Step 1 arms of the trial in approximately the proportions that were expected given the difference in numbers recruited to the WBA and UCA arms. Randomisation was stratified by site, and there was equal balance of participants in each of the Step 2 arms who had received usual care and active management advice. The proportion of participants being randomised to Step 2 was similar across all recruiting sites (Table 28). Cumulative target and actual recruitment is shown in Figure 10.

TABLE 28 - Number of patients recruited (% of Step 1 patients) by physical intervention and EDs

Cluster number		Advice n (% of Step 1)	Physiotherapy n (% of Step 1)	Not randomised n (% of Step 1)
UCA
1	University Hospitals Coventry and Warwickshire NHS Trust	33 (9%)	32 (9%)	309 (84%)
2	North Bristol NHS Trust	19 (8%)	17 (7%)	194 (85%)
3	Gloucestershire Hospitals NHS Trust	41 (10%)	42 (10%)	352 (82%)
4	Worcestershire Acute Hospitals NHS Trust	24 (8%)	26 (9%)	253 (84%)
5	Kettering General Hospital NHS FoundationTrust	16 (8%)	14 (7%)	183 (86%)
6	Buckinghamshire Hospitals NHS Trust	3 (5%)	5 (9%)	47 (85%)
WBA
7	Heart of England NHS Foundation Trust	51 (7%)	50 (7%)	616 (87%)
8	University Hospitals Birmingham NHS Foundation Trust	29 (6%)	31 (6%)	468 (89%)
9	Oxford Radcliffe Hospitals NHS Trust	23 (10%)	24 (11%)	180 (80%)
10	South Warwickshire General Hospitals NHS Trust	25 (10%)	25 (10%)	202 (80%)
11	Gwent Healthcare NHS Trust	14 (11%)	15 (12%)	101 (78%)
12	Countess of Chester Hospital NHS Foundation Trust	21 (5%)	19 (5%)	372 (90%)

FIGURE 10.

Step 2 recruitment versus target graph by advice intervention.

Characteristics of the population at randomisation are summarised in Table 29. The groups were well matched, with no baseline imbalances apart from a five-point difference in mean NDI scores between the groups [physiotherapy 44 (SD 16), advice 39 (SD 16)], indicating that participants randomised to physiotherapy had slightly greater neck disability. Analyses were therefore adjusted for this baseline imbalance.

TABLE 29 - Characteristics of Step 2 participants at randomisation

Collected at research clinic for Step 2.

Whiplash-associated disorder grades 0 and IV were not eligible for the trial.

Full-time employed, part-time employed, and self-employed.

Unemployed, full-time student, and retired/looking after home/inactive.

Mechanism of injury, location of pain, pain intensity, and medical history were collected at ED attendance. Gender, age, and ethnic group were collected on the 2-week follow-up questionnaire. Gender and age were also collected at the research clinic.

	Advice	Missing	Physiotherapy	Missing
Number randomised	299		300
Sex – male	115 (38%)	0	106 (35%)	0
Age in years, mean [SD]	40 [13]	0	40 [13]	0
Ethnic Group		16		20
White	229 (77%)		226 (75%)
Mixed	2 (67%)		3 (1%)
Indian	18 (6%)		19 (6.3%)
Pakistani	19 (6.4%)		19 (6.3%)
Bangladeshi	2 (67%)		1 (33%)
Black or Black British	10 (3.3%)		7 (2.3%)
Chinese or other	3 (1%)		5 (1.7%)
Mechanism of injury		2		1
Road traffic accident	284 (95%)		286 (95%)
Other	13 (4.3%)		13 (4.3%)
Location of pain		8		12
C-spine only	178 (60%)		178 (59%)
C-spine and other spinal area	60 (20%)		62 (21%)
Other spinal area only	10 (3%)		3 (1%)
Spinal + other area	27 (9%)		27 (9%)
Other area only	3 (1%)		6 (2%)
No pain	13 (4%)		12 (4%)
Pain intensity (/10), mean [SD]	5.4 [1.9]	69	5.6 [1.9]	91
History
Previous neck problems	36 (12%)	21	41 (14%)	19
Previous back problems	40 (13%)	56	43 (14%)	65
Neurological symptoms	26 (9%)	19	29 (10%)	15
WAD gradesa		0		0
0: No neck complaints or signsb	0 (0%)		0 (0%)
I: Complaints of pain, stiffness or tenderness, no physical signs	39 (13%)		45 (15%)
II: Complaint of pain, stiffness or tenderness, musculoskeletal signs	222 (74%)		220 (73%)
III: Complaint of pain, stiffness or tenderness, neurological signs	38 (13%)		35 (12%)
IV: Fracture/dislocationb	0 (0%)		0 (0%)
SF-12v1 scores, mean [SD]
Norm-based MCS	37 [12]	53	35 [12]	55
Norm-based PCS	36 [7]	53	36 [6.9]	55
Received any public funds	91 (30%)	17	101 (34%)	20
NDI (%), mean [SD]a	39 [16]	3	44 [16]	7
Employment		22		29
Working/earningc	225 (75%)		210 (70%)
Unpaid work	0 (0%)		2 (1%)
Not workingd	52 (17%)		59 (20%)

The flow of participants through the trial is summarised in Figure 11. There were 17 notifications of withdrawal from the trial (12 in the advice group and 5 in the physiotherapy group), and these participants were not followed-up at time points after their withdrawal (Table 30). Outcome data were available for 92% of participants at 4 months, 87% at 8 months and 80% at 12 months. There were no serious adverse events in Step 2.

FIGURE 11.

Step 2 CONSORT flowchart. (a) Two symptoms not appropriate, 15 not willing to participate; (b) 55 failed to attend, and 22 cancelled.

TABLE 30 - Reasons for withdrawals

Reasons	Advice	Physiotherapy	Total
Moved away	2	1	3
No longer interested	4	0	4
No reasons given	1	0	1
Recovered and no longer interested	1	0	1
Unhappy with trial	4	4	8
Total	12	5	17

Comparison of physiotherapy versus advice

For all outcomes, two series of analyses were performed; first, a complete case analysis, using all participants with data at each time point; and, second, a multiply imputed analysis to allow for the effects of missing data. Both analyses were adjusted for the effects of clustering in Step 1, NDI score at randomisation, and therapist. Estimates of the ICC for therapist effects are provided in Appendix 6. The ICC for therapists was consistent with only small proportions of variance (< 2%) being explained by therapists, and made minimal difference to the statistical models.

Primary outcome

The NDI scores for the physiotherapy group were on average 3.2 (95% CI –5.8 to 0.7) percentage points lower than those of the advice group at 4-month follow-up, but there was no evidence of a difference at 8 and 12 months. Physiotherapy therefore has a modest beneficial effect in the short term (see Table 35, Figures 12–14).

TABLE 35 - Neck Disability Index score (%)

Complete cases estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

Multiply imputed estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

	4 months	8 months	12 months
Advice
Mean (SD)	27.8 (17.4)	21.5 (16.7)	19.5 (17.0)
n (missing/total) (%)	5/255 (2%)	11/246 (4%)	3/238 (1%)
Physiotherapy
Mean (SD)	28.0 (17.9)	24.1 (18.4)	21.7 (18.4)
n (missing/total) (%)	11/252 (4%)	7/251 (3%)	6/241 (2%)
Treatment estimatea
∆ (95% CI)	–3.2 (–5.8 to –0.7)	–0.2 (–2.9 to 2.4)	–1.2 (–4.0 to 1.5)
Treatment estimateb
∆ (95% CI)	–3.7 (–6.1 to –1.3)	–1.0 (–3.6 to 1.6)	–2.0 (–4.6 to 0.6)

FIGURE 12.

Observed NDI scores (%) and 95% CIs by physical intervention.

FIGURE 13.

Unadjusted change in NDI from baseline research clinic for patients in Step 2 by individually-randomised treatment groups and month of follow-up.

FIGURE 14.

Multiply imputed estimate of physiotherapy effects on NDI by month of follow-up for patients in Step 2.

Secondary outcomes

There were no statistically or clinically significant differences in the treatment outcomes for the physical or mental health components of the SF-12v1, nor for acute whiplash injury and LWS (Tables 36–39).

TABLE 36 - Mental component score of SF-12v1

Complete cases estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

Multiply imputed estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

	4 months	8 months	12 months
Advice
Mean (SD)	45.9 (12.5)	47.1 (11.4)	48.8 (10.6)
n (missing/total) (%)	54/255 (21%)	54/246 (22%)	69/238 (29%)
Physiotherapy
Mean (SD)	46.3 (12.1)	46.2 (12.7)	47.5 (11.8)
n (missing/total) (%)	65/252 (26%)	60/251 (24%)	81/241 (34%)
Treatment estimatea
∆ (95% CI)	1.4 (–0.9 to 3.7)	–0.1 (–2.5 to 2.2)	0.1 (–2.3 to 2.4)
Treatment estimateb
∆ (95% CI)	1.3 (–0.9 to 3.5)	–0.3 (–2.6 to 2.0)	–0.0 (–2.2 to 2.1)

TABLE 37 - Physical component score of SF-12v1

Complete cases estimate adjusted for clustering, advice intervention, NDI score at Step 2 research clinic and therapists.

Multiply imputed estimate adjusted for clustering, advice intervention, NDI score at Step 2 research clinic and therapists.

	4 months	8 months	12 months
Advice
Mean (SD)	43.5 (9.7)	45.8 (9.5)	47.1 (9.9)
n (missing/total) (%)	54/255 (21%)	54/246 (22%)	69/238 (29%)
Physiotherapy
Mean (SD)	42.3 (9.2)	45.3 (9.8)	46.5 (10.2)
n (missing/total) (%)	65/252 (26%)	60/251 (24%)	81/241 (34%)
Treatment estimatea
∆ (95% CI)	–0.1 (–1.8 to 1.6)	0.5 (–1.3 to 2.3)	1.0 (–1.0 to 3.0)
Treatment estimateb
∆ (95% CI)	0.2 (–1.4 to 1.8)	0.2 (–1.5 to 2.0)	1.1 (–0.7 to 2.9)

TABLE 38 - Acute whiplash injury

Complete cases OR estimate adjusted for clustering, advice intervention, NDI score at Step 2 research clinic and therapists.

Multiply imputed OR estimate adjusted for clustering, advice intervention, NDI score at Step 2 research clinic and therapists.

	n (%) or odds ratio (95% CI)
Advice
Present	135 (47%)
Absent	120 (42%)
n (missing/total) (%)	32/287 (11%)
Physiotherapy
Present	135 (47%)
Absent	114 (40%)
n (missing/total) (%)	38/287 (13%)
Treatment estimatea
∆ (OR 95% CI)	0.81 (0.55 to 1.20)
Treatment estimateb
∆ (OR 95% CI)	0.76 (0.51 to 1.12)

TABLE 39 - Late whiplash syndrome

Complete cases OR estimate adjusted for clustering, advice intervention, and NDI score at Step 2 research clinic.

Multiply imputed OR estimate adjusted for clustering, advice intervention, and NDI score at Step 2 research clinic.

	n (%) or odds ratio (95% CI)
Advice
Present	86 (30%)
Absent	152 (53%)
n (missing/total) (%)	49/287 (17%)
Physiotherapy
Present	100 (49%)
Absent	141 (49%)
n (missing/total) (%)	46/287 (16%)
Treatment estimatea
∆ (OR 95% CI)	0.91 (0.59 to 1.39)
Treatment estimateb
∆ (OR 95% CI)	0.84 (0.56 to 1.28)

Physiotherapy had a beneficial effect on absences from work; participants randomised to physiotherapy treatment had on average a 41.4% reduction in work days lost over the first 4 months since their injury (95% CI –45.4% to –37.0%). This effect persisted to 8 and 12 months (Tables 40–42). We also performed an analysis of log-transformed data (using a bootstrap procedure to calculate CIs), to obtain an estimate of the absolute difference in work days lost (adjusted for clustering, advice intervention, NDI score at baseline and therapist). The results of this analysis were consistent with the Poisson regression. The estimated reduction in work days lost was 4.2 days (95% CI 1.1 to 7.9 days) (see Appendix 10).

TABLE 40 - Work days lost (cumulative) in Step 2 (Poisson regression)

Complete cases analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

Multiple imputation analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

	4 months	8 months	12 months
Advice
Mean (SD)	8.9 (22.9)	9.9 (25.3)	10.7 (26.2)
Missing/total	1/255 (0%)	0/246 (0%)	0/238 (0%)
Physiotherapy
Mean (SD)	6.3 (14.8)	8.2 (17.7)	9.2 (18.9)
Missing/total	0/252 (0%)	1/251 (0%)	1/241 (0%)
Treatment effect
Δ (95% CI)a	–41.4 (–45.4 –37.0)	–33.3 (–37.5 to –28.9)	–17.5 (–22.5 to –12.0)
Δ (95% CI)b	–43.5 (–50.0 to –36.1)	–31.2 (–41.3 to –19.5)	–15.13 (–27.0 to –1.4)

TABLE 41 - Work days lost (cumulative) in Step 2 additionally adjusted for not working at baseline (Poisson regression)

Complete cases analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

Multiple imputation analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

	4 months	8 months	12 months
Δ (95% CI)a	–39.3 (–43.6 to –34.5)	–33.6 (–37.9 to –29.0)	–8.3 (–14.1 to –2.1)
Δ (95% CI)b	–37.0 (–44.4 to –28.6)	–26.7 (–37.1 to –14.6)	–8.21 (–21.1 to 6.8)

TABLE 42 - Work days lost (cumulative) in Step 2 restricted to those who worked at baseline (Poisson regression)

Complete cases analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

Multiple imputation analysis; percentage difference adjusted for therapist clustering within ED, The Whiplash Book advice, NDI score at baseline and time from ED attendance.

	4 months	8 months	12 months
Advice
Mean (SD)	11.1 (25.4)	12.3 (28.1)	13.3 (29.1)
Missing/total	0/202 (0%)	0/192 (0%)	0/184 (0%)
Physiotherapy
Mean (SD)	7.4 (15.2)	9.4 (18.3)	10.6 (19.7)
Missing/total	0/189 (0%)	1/185 (1%)	1/180 (1%)
Treatment effect
Δ (95% CI)a	–42.6 (–46.8 to –38.0)	–36.6 (–40.9 to –32.1)	–13.5 (–19.2 to –7.5)
Δ (95% CI)b	–41.0 (–49.6 to –30.8)	–31.6 (–42.7 to –18.3)	–13.4 (–26.9 to 2.4)

Subgroup analyses: injury severity, previous neck pain and early psychological response

There were no statistically significant interactions between the NDI and severity of initial injury, psychological reaction to the injury or previous neck pain (Tables 43–45). Estimates (mean and 95% CIs) of all terms tested in the NDI subgroup analyses are given in Appendix 7.

TABLE 43 - Subgroup analysis of NDI by initial injury severity

Complete cases estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

Multiply imputed estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Severe symptoms
Advice	28.7 (17.5) [216]	22.5 (16.7) [203]	20.3 (17.2) [204]
Physiotherapy	29.5 (18.1) [202]	25.2 (19.1) [205]	23.0 (19.0) [197]
Milder symptoms
Advice	21.9 (16.0) [34]	15.5 (15.3) [32]	14.2 (14.6) [31]
Physiotherapy	20.0 (14.6) [39]	18.4 (13.3) [39]	15.1 (13.0) [38]
p-value for interactiona	0.23	0.96	0.35
p-value for interactionb	0.24	0.62	0.42

TABLE 44 - Subgroup analysis of NDI by early psychological response

Complete cases estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

Multiply imputed estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Early psychological response present
Advice	29.6 (17.8) [178]	23.1 (17.3) [166]	21.1 (17.8) [160]
Physiotherapy	30.6 (18.9) [176]	26.7 (19.7) [177]	24.1 (19.4) [173]
Early psychological response absent
Advice	22.5 (13.9) [68]	17.0 (13.4) [66]	15.7 (14.2) [71]
Physiotherapy	20.5 (12.0) [55]	17.6 (11.4) [57]	15.1 (13.0) [55]
p-value for interactiona	0.64	0.70	0.72
p-value for interactionb	0.66	0.86	0.45

TABLE 45 - Subgroup analyses of NDI by presence of previous neck pain

Complete cases estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

Multiply imputed estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Had previous neck pain
Advice	31.0 (19.8) [21]	26.7 (16.6) [23]	26.5 (17.0) [24]
Physiotherapy	34.4 (19.8) [19]	32.8 (19.1) [18]	28.3 (20.8) [14]
No previous neck pain
Advice	27.3 (17.1) [223]	21.0 (16.7) [207]	18.6 (16.8) [205]
Physiotherapy	27.3 (17.7) [215]	23.6 (18.2) [219]	21.5 (18.4) [214]
p-value for interactiona	0.96	0.72	0.41
p-value for interactionb	0.84	0.81	0.61

Compensation pursued

As in Step 1, we performed an exploratory analysis comparing the outcomes of those who pursued a compensation claim versus those who did not. This was limited in power by the fact that only 13 participants reported at 12 months that they had not pursued a claim for compensation.

Among participants who pursued compensation, those randomised to physiotherapy tended to have slightly higher NDI scores at all time points. However, among those who did not pursue compensation, participants randomised to physiotherapy had considerably lower NDI scores, indicating less neck injury-related disability (Table 46). The interaction between physiotherapy treatment and compensation approached statistical significance (p = 0.06) at 12-month follow-up for complete cases and multiply imputed models.

TABLE 46 - Subgroup analysis of NDI by whether or not a compensation claim was pursued

Complete cases estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

Multiply imputed estimate adjusted for clustering, advice intervention NDI score at Step 2 research clinic and therapists.

	4 months Mean (SD) [n]	8 months Mean (SD) [n]	12 months Mean (SD) [n]
Pursued compensation
Advice	26.7 (17.6) [210]	20.3 (16.0) [206]	19.3 (17.2) [227]
Physiotherapy	27.7 (17.5) [203]	24.1 (18.2) [208]	22.0 (18.4) [221]
Did not pursue compensation
Advice	34.0 (17.0) [2]	25.2 (10.0) [5]	28.8 (7.3) [5]
Physiotherapy	12.7 (13.3) [6]	20.0 (21.1) [8]	11.0 (14.7) [8]
p-value for interactiona	0.27	0.77	0.06
p-value for interactionb	0.53	0.68	0.06

The interaction term between SF-12v1 physical component score and compensation was also significant at 12-month follow-up for complete cases model, but was no longer significant in the multiply imputed model. Given the small number of participants who did not pursue compensation, it is difficult to assess the possible significance of these results.

There were no other suggestions of any interaction between any the treatment group and compensation for any outcome.

Additional analyses: the effectiveness of physiotherapy interventions in combination with different advice interventions

We investigated the potential interactions between the two physiotherapy interventions provided in Step 2 and the two advice interventions provided in Step 1, in those participants randomised to Step 2. The randomisation procedure resulted in an even distribution of characteristics within each of the four possible treatment combinations (UCA/physiotherapy advice session; UCA/physiotherapy package; active management advice/ physiotherapy advice session; active management advice/physiotherapy package; shown in Table 47).

Mechanism of injury, location of pain, pain intensity, and medical history were collected at ED attendance. Gender, age, and ethnic group were collected via the 2-week follow-up questionnaire in Step 1.

Tables 48–55 give the estimate of treatment effectiveness for physiotherapy and The Whiplash Book elements separately for the NDI and other outcomes, and the interaction between them. There was no suggestion of an interaction between the treatments provided in Step 1 and those provided in Step 2.

Additional analyses: recovery in participants not randomised to Step 2

We examined recovery in the people who participated in Step 1 but were not randomised to Step 2 (Table 56). Figure 15 demonstrates that overall symptoms were milder than those within Step 2, but that recovery was not complete at 12 months.

FIGURE 15.

Neck Disability Index (%) for those who did not enter Step 2 by advice intervention.

Introduction

The aim of the qualitative study was to gain the participant’s perspective on the experience of having a whiplash injury, and the experience of treatment and recovery within the context of MINT. Despite the recognised importance of qualitative data to understanding patient experiences, there are very few qualitative studies in the field of WADs. 68

Method

Semi-structured interviews were conducted with 20 participants after they had received their allocated treatments. All people interviewed were participants in Step 1 and 2 of the trial.

Participant recruitment

While attending the clinic for entry to Step 2 of the trial, 25 people were invited to participate, given an information sheet detailing the purpose and content of the interviews, and advised that a researcher (EMW) would contact them on completion of their treatment. Of these, 20 were contacted and completed the interview and one was contacted and arranged an interview but failed to attend (Figure 16).

FIGURE 16.

Participant recruitment to the interview study.

Results

Step 2 interventions

Conclusion

This study has explored the experience of treatment following a whiplash injury, particularly how the different components of the treatment were perceived and used by participants and how they impacted on recovery.

In the narratives recorded in this small group of participants, the recall of the content of verbal advice given in the EDs was similar regardless of the arm of the trial. Although generally of high quality, the advice given was not always consistent with the concept of active management as a few participants reported being told to rest. The interviews were conducted after all treatments had been completed, and it is probable that the recall of specific details of treatments given is limited by recall bias. However, given the long recall period it is perhaps surprising that people do remember at least some of the content of the ED consultation. Messages regarding physical activity and rest appear to be retained over time. Although it would be important to understand the context in which instructions to rest had been made, the indication is that the educational packages given to clinicians in EDs may have been insufficient to ensure that all clinicians had embraced the core components of the active management strategy.

The interviews suggest that The Whiplash Book is well received by patients. Participants receiving The Whiplash Book and active management advice described greater engagement with the written advice, than those receiving the usual care leaflet. A higher proportion of participants receiving The Whiplash Book initiated the recommended exercises promptly. The narratives suggest that The Whiplash Book gives added value over the usual care leaflet. However, we need to be cautious in drawing conclusions. All participants in the interview study were participants in the second stage of the trial, and, as such, the advice encapsulated in the original ED consultation was reinforced during the follow-up treatment. The added value of The Whiplash Book may be, at least in part, that it provides an adequate basis for consistent reinforcement by health professionals who may see people later in the course of their recovery. However, on its own The Whiplash Book seems to have enabled more people to start on exercises for their neck than the usual care leaflets.

Despite receiving written advice, some participants expressed hesitation about doing exercises without further guidance or reassurance. This hesitation was in part because participants wanted reassurance that they were doing the exercises in the right way, but some participants expressed a fear of making themselves worse. The hesitant behaviour was found among participants in the usual care and active management advice arms. Fear of damage is known to lead to activity restriction in other musculoskeletal problems, and is hypothesised to be an important mediator on the path to chronic symptoms. 72 For individuals where fear avoidance plays an important role in their reaction to injury and recovery, The Whiplash Book alone does not seem to be sufficient to allay their fear. Participants reported using The Whiplash Book selectively, often only using the exercise component. Improving the accessibility and format of the other sections of the booklet may improve effectiveness. Providing advice about what to do if you cannot get started on activity and movement may prove a useful addition to the text.

All participants reported that a follow-on session or sessions with a physiotherapist were important in allaying their hesitation to exercise, and enabling them to start exercise.

The single physiotherapy advice session appeared sufficient to many participants. Great value was placed on rapid access to the service. Some participants felt they would have benefited from a follow-up session, or at least have this as an option. For participants it was important that the physiotherapist assessed them as individuals and helped them understand how to undertake the exercises, including demonstrating the exercises.

Those attending the physiotherapy package also highlighted guidance, reassurance, and continuing exercises as helpful. Some participants appreciated manual therapy. For most participants, the physiotherapy package was considered sufficient. More flexible appointment times may have made it easier for some participants to attend, and allows people to be able to continue with important commitments such as work. Although it is common practice for a package of physiotherapy to be delivered by the same physiotherapist, future research should examine whether or not this continuity in delivery is essential.

Some participants sought advice or treatment for their whiplash beyond the trial.

There are a number of limitations and caveats to note. Qualitative research usually involves in depth analysis of small samples, and it is not possible (nor an intended aim of the method) to generalise the findings to a wider population. We undertook purposive sampling to try to maximise the representativeness of the sample in terms of injury severity and gender, and of the overall clinical trial sample. We interviewed 20 participants, with data saturation sufficient to develop a meaningful and internally consistent thematic framework.

There are some limitations specific to this study. The length of time between attending the ED and having the qualitative interview may have affected recall, particularly of treatments delivered in the first step of the trial. Also we need to highlight that remembering advice does not equate to changing exercise, illness and pain behaviours. Some of the participants had been registered in to the second step of the trial by the same researcher who conducted the interviews, and hence may have been reluctant to provide criticism of the treatments provided.

In conclusion, these interviews suggest – that The Whiplash Book can give added value over the usual care provided in the ED. However, as intentions do not necessarily translate into trying exercises, and the non-exercise components of the book are overlooked, The Whiplash Book may not affect clinically important end points. Both the physiotherapy advice session and the physiotherapy package were beneficial to participants. Consistent reinforcement of the same messages about how to manage a whiplash injury was helpful for many participants particularly when a therapist engaged with the individual and tailored the messages to each individual.

Introduction

This section reports the cost-effectiveness analysis of the stepped care approach to managing neck injuries including education and advice based on The Whiplash Book at the initial contact, compared with usual care in EDs, and, in patients with persisting symptoms, the supplementation of reinforcement of advice with physiotherapy treatment, compared with reinforcement of advice only.

We assessed the cost–utility of advice based on The Whiplash Book and of physiotherapy. The cost–utility analysis compared The Whiplash Book with usual care and separately, physiotherapy with reinforcement of advice, reflecting the study design. The primary analysis was a within-trial cost–utility, expressed as the cost–utility ratio, the cost of gaining one additional quality-adjusted life-year (QALY) with each of the two interventions considered.

The cost–utility was assessed over a time frame of 12 months. Costs were expressed in UK pounds (£) actualised to current prices using the Office for National Statistics Retail Price Index. 73 2009 was the base year of the analysis. Costs and QALYs were not discounted as the time horizon of the analysis was 1 year.

We calculated the cost–utility from both the UK NHS perspective and a general health-care perspective including all private health care and related costs supported by participants.

Data

Analysis

The base-case analysis included the sample of all participants who returned at least one resource use questionnaire at 4 months. Missing data were imputed only in a limited number of cases. For individuals who had missing NHS costs, data were imputed if the respondent had indicated the use of a resource but not the quantity. Data were not imputed for self-funded items as the patients declared the total amount disbursed; therefore, the cost of each type of private service consumed was not available.

The cost–utility analysis was conducted using data from cases that had completed at least one follow-up questionnaire. For patients who failed to return at least one questionnaire, data were not imputed, as the inverse-weighted analysis (illustrated below) allows adjusting for the total time of follow-up for each patient. Furthermore, the number of missing resource-use and cost data was low.

We also conducted a descriptive analysis of cost and resource-use data to provide cost data in reusable form. To ensure that the resource use and costs were based on a homogeneous time period, the descriptive analysis was conducted on complete cases only, including cases that had returned resource consumption data over a 12-month period.

Quality-adjusted life-years

The EQ-5D data obtained from participants questionnaires were converted into utility weights using an econometric model developed from a survey of time trade-off valuations of health states from members of the British general public. 31 Utility weights are a measure of the relative value of a health state compared with perfect health and are between 0 and 1. These extremes indicate the utility associated with death and perfect health, respectively.

Quality-adjusted life-years were then calculated for each participant using the area under the curve, which is the weighted sum of the utility at baseline and at 4, 8 and 12 months, and the time spent in each period expressed in years.

The trial did not collect utility data at baseline, because participants were enrolled at EDs with minimal data collection. Utility data were collected in the 2-week, 4-month, 8-month and 12-month follow-up questionnaires. Therefore, we imputed utility values at baseline using utility data from the general population derived from the Health Survey for England,79 matching values by age. This is justified as the population in this trial is made up by individuals from the general population who had an accident, but were otherwise no different from the general public. Baseline utility values were applied from the day of randomisation to 2 weeks, whereas data on utility collected in the trial were used thereafter.

Descriptive summaries for utility weights were calculated and the difference in weights between interventions and controls was tested using t-tests for differences in mean.

We explored the determinants of QALYs using a regression-based approach, including clinical and behavioural prognostic factors, such as a range of clinical indicators of severity of neck injury and neck pain, age, gender, psychological problems, utility at baseline and whether or not the participant sought compensation for neck injury.

Cost of care

Total cost of care was computed for each participant, multiplying the number and type of resources used by their unit costs.

The total cost of care was computed from the NHS and from the general health-care perspectives, including the cost components indicated previously (see Chapter 6, Data). The broad classes of costs included in the NHS perspective were the cost of consultations, admissions, physical and other therapies, drugs and diagnostic tests. The health-care perspective included, in addition to NHS costs, the costs of consultations, admissions, therapies, drugs, diagnostics and equipment privately purchased.

We conducted a descriptive analysis of mean costs using a complete case analysis at 4, 8 and 12 months. This includes patients who returned all questionnaires and patients whose data were collected on follow-up telephone calls if non-respondent and were asked to provide NHS and private resource consumption data.

Mean resources and costs were tabulated for each type of health-care resource separately for Step 1 and Step 2 of the study. Mean differences in costs by groups were tested for both steps of the study, using a t-test of difference in means.

As the cost of care is not symmetrically distributed, standard errors were also constructed using bootstrap procedures, sampling 5000 random samples from the original study data and calculating the mean at each sample. The 95% confidence bounds were computed from the empirical distribution of the means obtained from the bootstrap procedure.

We explored the determinants of costs using the same prognostic factors used in the analysis of QALYs, indicators of severity of neck injury and neck pain, age, gender, psychological problems, utility at baseline and whether or not the participant sought compensation for neck injury. This analysis aimed to establish the significant prognostic factors included in the incremental cost-effectiveness analysis. We analysed the significance of prognostic factors using ANOVA.

Incremental analysis

The incremental cost–utility ratio (ICUR) was calculated as the ratio between the difference in the mean cost of care and the difference in mean QALYs. The incremental cost–utility was calculated separately for the Step 1 and Step 2 comparisons, The Whiplash Book compared with standard care and physiotherapy compared with reinforcement of advice.

Incremental QALYs and costs were calculated using a regression approach. Mean costs and mean QALYs were estimated in a regression including a term for treatment, and controlling for length of follow-up and baseline utility as continuous factors, as these factors are correlated with total QALYs and total costs. The regression was also controlled for prognostic factors that emerged from the descriptive analysis of costs and QALYs.

The difference in mean costs and mean QALY was the coefficient of the treatment term in the regression. A positive coefficient indicated that treatment was an independent predictor of increased costs and of health gain, (expressed in QALYs); therefore, the ICUR was computed as the ratio of the coefficient of costs and QALYs obtained from the regressions.

Mean QALYs were estimated as the inverse probability weighted mean of the total quality-adjusted survival for each participant in the four groups. The use of inverse probability weighting is justified to account for censoring of participants data owing to loss of follow-up and as the accrual of costs and QALYs are believed to be correlated with the duration of participation into the study. Therefore, the accrual of QALYs may differ between patients who were followed up for the entire duration of the trial and those that were lost to follow-up at earlier dates. In this analysis we applied the inverse-weight adjustment to the estimation of QALYs only.

Furthermore, we used the regression approach to control for factors that may be important in the prediction of costs and quality of life (QoL). The estimation of mean costs and QALYs should be adjusted a priori by utility at baseline. This is because the value of utility at baseline correlates with the total QALY as it enters its calculation directly. Any randomised comparison between treatments is likely to show a difference in utility by group at baseline. Any imbalance in the utility at baseline, regardless of statistical significance, should always be regarded as a source of bias in the computation of the total QALY difference between the two groups. In addition, such analysis allows for the incorporation of adjustments for control factors depending on the results of the cost and QALY analysis, which may reveal additional significant predictors of costs and QALYs.

To explore the uncertainty in the cost–utility analysis, we used a non-parametric bootstrap approach, sampling 5000 random samples for costs and QALYs from the original study. A group of equal number of cases than the original study groups was sampled for each bootstrap iteration and the incremental costs and QALYs were calculated using a regression.

Incremental costs and QALYs were plotted on the cost-effectiveness plane, which shows the joint distribution of the difference in costs and the difference in QALYs generated by the bootstrap procedure. The cost-effectiveness plane is divided in four quadrants, each of them representing a potential combination of incremental costs and QALYs:

1. The active intervention (The Whiplash Book or physiotherapy) is more effective than the comparator (usual care or reinforcement of advice) but also more expensive. This combination identifies points on quadrant I of the cost-effectiveness plane. In this case, the intervention is deemed cost-effective if the incremental cost–utility ratio falls below the societal decision maker’s willingness to pay (WTP) the monetary value that the decision maker is willing to pay for one additional QALY. The societal decision maker’s WTP has proven empirically difficult to set, although it has been suggested that, for the UK, the societal cost-effectiveness threshold could be at around £30,000. Therefore, the cost–utility was presented over the range £0–50,000. The decision-maker’s cost-effectiveness threshold is also known with the symbol λ.

2. The active intervention is both less costly and less effective than the comparator (points on quadrant III). This combination indicates that the usual care or reinforcement of advice is cost-effective and therefore the comparator should be chosen if the ICUR is lower than the decision maker’s cost-effectiveness threshold.

3. The Whiplash Book or physiotherapy is less expensive, yet more effective than usual care or reinforcement of advice (quadrant IV). Then the interventions are ‘dominant’ therefore should be always chosen.

4. The Whiplash Book or physiotherapy is more expensive and less effective (quadrant II). Therefore, the interventions are ‘dominated’ as there is no cost-effectiveness threshold at which the decision maker would be willing to pay.

The cost-effectiveness plane provides a visual illustration of the extent of cost-effectiveness or dominance combinations and the extent of cost–effective combinations for The Whiplash Book or physiotherapy.

Because of the presence of dominance combinations, the variability of the incremental cost–utility ratio cannot be represented with confidence intervals. Therefore, we investigated the uncertainty around the ICUR calculating the net benefit (NB) statistic,

which is a summary statistic for the net value of The Whiplash Book or of physiotherapy expressed as the monetary value of the incremental gain, weighted by the WTP of the decision maker, and net of the cost of the intervention. Therefore, an intervention is cost-effective when the total monetary value of health gain is higher than its costs, hence when the net benefit is positive.

From the incremental costs and QALYs obtained from the bootstrap sample, the net benefit was calculated for a range of cost-effectiveness threshold values. The empirical proportion of positive net benefit was estimated and plotted on a graph, the cost-effectiveness acceptability curves (CEACs), a representation of the probability that each intervention is cost-effective given the range of WTP of the decision-maker. CEACs were presented for both The Whiplash Book and physiotherapy.

The cost–utility analysis was reported presenting the results of the regression models used to estimate the incremental QALYs and the incremental costs. To explore the impact of methodological assumptions, we report the results of the cost–utility analysis with and without the inverse probability weight adjustment applied in the regression models used to estimate incremental costs and QALYs. Furthermore, to assess the performance of the bootstrap procedure, we also report the results of the regressions estimated using the bootstrap procedure and compare them with the results of the exact regression models.

For each comparison, The Whiplash Book versus usual care and physiotherapy versus reinforcement of advice, we then report the CEACs illustrating the probability of cost-effectiveness over a range of decision makers’ WTP between £0 and £50,000.

Results

Incremental cost-effectiveness

Incremental cost-effectiveness: NHS

The ICER was obtained from the regressions of costs and QALYs for the two steps of the study.

The Whiplash Book was associated with higher costs and lower QALYs and therefore was dominated by usual care. Similarly, physiotherapy was both more expensive and less effective than reinforcement of advice (Table 77).

TABLE 77 - Incremental cost-effectiveness ratios estimated using the non-inverse-weighted method

	Incremental costs (£)	Incremental QALYs	ICER
Non-inverse weight-adjusted (deterministic results)
Step 1
The Whiplash Book vs usual care	27.95	–0.00298	Dominated
Step 2
Physiotherapy vs reinforcement of advice	58.36	–0.0111	Dominated
Non-inverse weight-adjusted (probabilistic, bootstrapped expectation)
Step 1
The Whiplash Book vs usual care	27.54	–0.0031	Dominated
Step 2
Physiotherapy vs reinforcement of advice	59.27	–0.0112	Dominated

These results were invariant with respect to the methods used to estimate costs and QALYs. In the bootstrap analysis, both The Whiplash Book in Step 1 and physiotherapy in Step 2 are dominated at any level of WTP. The inverse weighting also did not have an impact on the results (Tables 77 and 78).

TABLE 78 - Incremental cost-effectiveness ratios using the inverse-weights method

	Incremental costs (£)	Incremental QALYs	ICER
Inverse weight-adjusted (deterministic results)
Step 1
The Whiplash Book vs usual care	27.95	–0.0046	Dominated
Step 2
Physiotherapy vs reinforcement of advice	58.36	–0.0115	Dominated
Inverse weight-adjusted (probabilistic, bootstrapped expectation)
Step 1
The Whiplash Book vs usual care	27.54	–0.0041	Dominated
Step 2
Physiotherapy vs reinforcement of advice	59.27	–0.0111	Dominated

The CEACs (Figure 17) for Step 1 and Step 2 show that both interventions at Step 1 (The Whiplash Book) and Step 2 (physiotherapy) are dominated by the comparators at any threshold of cost-effectiveness used by decision makers.

FIGURE 17.

Cost-effectiveness acceptability curves for Step 1 and Step 2: NHS perspective (inverse probability weighted model).

Incremental cost-effectiveness: general health-care perspective

The results of the incremental cost-effectiveness analysis from the viewpoint of the general health-care sector are similar to those of the NHS. The Whiplash Book is not cost-effective compared with usual care (Table 79). The intervention reduces the total QALYs slightly; however, the reduction in costs is not sufficient to compensate for the loss of health outcomes and therefore usual care remains cost-effective at a ICER of approximately £7000. Physiotherapy remains dominated as it remains associated with an increase in costs and a decrease in health outcomes. Again these results are robust to the analytical methods adopted in the cost-effectiveness (Tables 79 and 80).

TABLE 79 - Incremental cost-effectiveness ratios estimated using the non-inverse-weighted method

	Incremental costs (£)	Incremental QALYs	ICER
Non-inverse weight-adjusted (deterministic results)
Step 1
The Whiplash Book vs usual care	–21.93	–0.0030	£7359 Usual care is cost-effective
Step 2
Physiotherapy vs reinforcement of advice	48.29	–0.0111	Dominated
Non-inverse weight-adjusted (probabilistic, bootstrapped expectation)
Step 1
The Whiplash Book vs usual care	–20.62	–0.0031	£6652 Usual care is cost-effective
Step 2
Physiotherapy vs reinforcement of advice	53.54	–0.0112	Dominated

TABLE 80 - Incremental cost-effectiveness ratios using the inverse weighted method

	Incremental costs (£)	Incremental QALYs	ICER
Inverse weight-adjusted (deterministic results)
Step 1
The Whiplash Book vs usual care	–21.93	–0.0046	£4767 Usual care is cost-effective
Step 2
Physiotherapy vs reinforcement of advice	48.29	–0.0115	Dominated
Inverse weight-adjusted (probabilistic, bootstrapped expectation)
Step 1
The Whiplash Book vs usual care	–20.62	–0.0041	£5029 Usual care is cost-effective
Step 2
Physiotherapy vs reinforcement of advice	53.54	–0.0111	Dominated

In the bootstrap analysis from the general health-care perspective, usual care is cost-effective at generally accepted cost-effectiveness thresholds. Physiotherapy remains dominated at any level of WTP. This is reflected in the CEACs for Step 1 and Step 2 (see Figure 18).

FIGURE 18.

Cost-effectiveness acceptability curves for Step 1 and Step 2: general health-care perspective (inverse probability weighted model).

Aims and overview of the trial findings

In 1995, the QTF on whiplash injuries and disorders recognised the need for a definitive trial of treatments commonly used in whiplash injury, and, in particular, for a trial of physiotherapy. This need was reiterated by the British Bone and Joint Decade Task Force on whiplash injury in 2008, which, on reviewing the accumulating evidence base, concluded that progress in terms of large, definitive trials had been poor, and fundamental questions about the clinical management of whiplash remained. 80

MINT is the largest trial of treatments for acute whiplash injury completed to date. It was designed to provide a definitive answer to several questions about the clinical management of acute WAD. The sample size was large, to enable detection of small but potentially worthwhile clinical benefits and to allow a robust economic evaluation alongside the trial. The trial was also designed to track and evaluate a two-step process of delivering treatment which is observed frequently in clinical practice, that being the provision of a ‘light touch’ treatment of advice in the first instance, followed by a more resource-intensive second step of treatment to those who are struggling to recover. 81

The first step of the trial examined whether or not an enhanced advice and active management strategy, supplemented by a relatively inexpensive and bespoke advice booklet, was more effective than usual care. Overall, it was not. The second step of the trial examined two approaches to providing additional physiotherapy: a single session of advice or a typical package of physiotherapy allowing for up to six sessions of physiotherapy treatment. Although there was some small benefit of the additional physiotherapy package at 4 months after the injury, in the longer term this made no difference to the clinical outcome, and importantly, from a health-care perspective, the small clinical gain was not worth the additional cost at current accepted levels of WTP. We conclude, overall, that for those individuals who perceive they are struggling with recovery, a single session of advice from a physiotherapist is sufficient. Our design does not allow us to draw a direct comparison of no follow-up treatment. However, the qualitative data collected suggest strongly that the advice session provides much needed reinforcement and encouragement to engage with doing exercises and taking a positive approach to recovery. Overall, the best combination of treatment was UCA followed by a single physiotherapy advice session for those struggling. The pattern of clinical response was consistent across similar outcomes.

Internal validity and methodological limitations

The trial was powered to detect a moderate effect size (0.375), assuming a pooled SD of eight absolute points (16 percentage points). Overall, the assumptions underlying the sample estimate appear correct. For Step 1 of the trial, the pooled SD at each of the follow-up points was within our anticipated range. For Step 2 of the trial, the pooled SD at baseline was within the anticipated range, although it increased slightly over the follow-up time points. We estimated an ICC coefficient of 0.02 and an average of 350 patients per centre, giving an inflation factor of 7.98.

In the early phases of the study, we recognised that we would not achieve the desired sample size, and increased the number of clusters recruiting into the trial. This had the overall effect of reducing the sample size. At the close of the trial, the point estimate for the ICC suggested a much smaller clustering. This, combined with a smaller loss to follow-up than anticipated, means that MINT had good power to detect the differences we originally specified.

Acute injury trials present a particular difficulty in being able to estimate accurately the pre-injury status. A number of options exist – to collect the data retrospectively (which is subject to recall bias) or to accept the premise that randomisation will result in well-matched groups. Within the constraints of a large pragmatic trial in which treatments and trial procedures are being implemented alongside the day-to-day operations of EDs, we did not have the option to collect an extensive data set. Pre-treatment WAD grades were collected and a four-point indicator of HRQoL.

Randomisation resulted in two groups that were well matched in the first step of the trial. Although the ethnic mix of the sample was representative of the UK population, there were slightly more participants of Pakistani origin in one arm of the trial than in the other. Sensitivity analyses concluded that this small imbalance made no qualitative or quantitative difference to the overall estimates of treatment effect. In the second step of the trial, there was a difference in the NDI scores collected at the research clinic. We have examined our procedures carefully. The risk of allocation subversion is small, as we used a quality-assured telephone randomisation system. We stratified by site, but did not use blocking, thus minimising the chance that recruitment staff could anticipate and hold patients for favourable allocations. Allocation lists have been checked for date and sequence, and all found to be normal. The most likely explanation is that these differences have arisen by chance. Overall, the difference will not have affected our overall estimate of treatment effect; the difference was accounted for in the analysis.

Although we did not account for clustering attributable to therapist effects in the original sample size estimation, the final models of treatment effect were adjusted for these effects. The ICC for therapist effects was of the order of 0.02 for the primary outcome, indicating a minimal amount of intertherapist variability in the intervention effectiveness. In comparison with other studies, the therapist effects are small, and carry most influence at the time points nearest to the delivery of treatment.

Loss to follow-up was lower than expected for a trial of an acute injury. Previous pragmatic studies of acute injury management in EDs have reported loss to follow-up of ≥ 30% at 9 months. 82 We used a system of repeat questionnaires, supplemented by a telephone interview, to collect the core primary outcomes where postal response could not be elicited. Telephone data collection worked well, although we were unable to collect as much information and the process required substantial dedication from the trial team, including out-of-hours working. Although there was a tendency for younger men to respond less often, our estimates of the treatment effectiveness appear insensitive to missing data.

Analytically the greatest challenge posed by the trial was the lack of EQ-5D estimate prior to initiation of treatment in the ED, and prior to the injury in the first step of the trial. We addressed this by imputing a value based on the age and gender expected norms for each individual. This inevitably introduces some underestimation of variance in the EQ-5D, and an assumption that the population we recruited from is broadly reflective of the age- and gender-matched population. Overall, however, these assumptions and shortfalls are small, and we are confident in the conclusions of the economic analysis for Step 1 of the trial. Baseline values were available for Step 2 of the trial.

External validity and generalisability of the findings

Overall we believe the generalisability of the trial to be good, with good representation of the injury grades usually encountered in clinical practice, a wide range of hospitals and substantial numbers of participants. We recruited 15 EDs, comprising large teaching hospitals, foundation hospitals and district general hospitals. The EDs served a mix of urban and semi-rural catchment areas. The services available in these hospitals reflected the normal services in UK hospitals, with none having specific specialist services applicable to whiplash injuries. The level of training provided in WAD management was consistent with the best that could be achieved under the constraints of routine clinical practice. The population recruited was reflective of the expected incidence of different injury severity within the UK, and was similar in terms of age and gender to previous UK series. The recovery trajectory was consistent with published rates of recovery from smaller studies based on the UK population where reports of persistent symptoms at least 1 year post injury ranged from 16% to 48% of participants. 21,22,83–85 In MINT, approximately 18% of participants in Step 1 and 32% in Step 2 had persistent symptoms at 12 months (LWS). Forty-eight per cent of people had recovered at 4 months in Step 1 and 41% in Step 2.

The wide, pragmatic inclusion criteria have allowed us to investigate the effects in pre-specified subgroups of patients. Subgroup effects should be analysed through formal tests of statistical interaction and be pre-specified to avoid potentially spurious results. 34 We adopted both these approaches. The sample size allowed for detection of large subgroup effects (approximately double the size of the main effect), provided imbalance between the subgroups was no greater than 70 : 30. We investigated three potential subgroup effects: (1) severity of the index injury; (2) adverse psychological reactions to the injury; and (3) neck pain prior to the injury. We cannot draw any conclusions relating to neck pain prior to the injury because the subsample of people with prior neck pain is too small (< 7%). The distribution of other pre-specified subgroups was within the acceptable range within the first step of the trial. We can conclude that there were no substantial interactions between the psychological factors or injury severity and the provision of advice. In the second step of the trial there was a statistically significant interaction between symptom severity and treatment allocation, suggesting that the physiotherapy package is more effective in people with severe symptoms. However, this should be interpreted with caution as there were relatively few cases with mild symptoms in the second step of the trial.

Interpretation and implications for clinical practice and policy

Overall, there was no benefit to training EDs in an active management strategy and supplementing this advice with The Whiplash Book. The most probable explanation is that active management (as promoted by The Whiplash Book) is no more effective (or ineffective) in encouraging people to return to normal activities or to initiate exercise, and hence influence clinical outcome.

Although the extended physiotherapy package resulted in some short-term symptomatic relief in neck disability (or more rapid restoration of function), this additional treatment had no impact on generic HRQoL. This is perhaps surprising, and there are several potential explanations. First is that injury has minimal impact on HRQoL. We used two measures of generic HRQoL in this trial, the EQ-5D and the SF-12. The injury is isolated in its nature, and is unlikely to substantially impact on the locomotor system, which is an important component of measures of generic HRQoL. The injury is painful, and our data suggest that it affects fatigue, sleep, driving and reading predominantly – these domains do not feature strongly in either the SF-12 or EQ-5D. The conclusion of the health economic analysis, which is based on currently accepted methods of using the EQ-5D to facilitate comparison of the benefits across conditions, is that the cost of the additional treatment dominates the benefits. Hence, despite some short-term symptomatic relief, we do not recommend an extended physiotherapy package as a cost-effective treatment for routine clinical practice. The single supplementary advice session with the physiotherapist does appear beneficial in encouraging people to exercise and undertake self-management activities, but this conclusion is based on the qualitative component of the study, not any quantitative comparison.

There are a number of other potential reasons why the extended package is less effective than anticipated. This was a pragmatic trial, and the treatment effect may have been diluted by participants seeking physiotherapy or similar treatment elsewhere. Self-reported resource use suggests that, overall, people randomised to the additional physiotherapy package did receive extra treatment, and examination of the treatment records suggests that the therapists were broadly compliant with the treatment protocol.

As with all trials of complex interventions, it was necessary to select the components of the intervention to be evaluated from a range of possible treatments. We selected the treatments that are most commonly applied in contemporary clinical practice, and have some evidence supporting their effectiveness. The treatment package contained exercise, manual therapy, advice and reassurance, all of which have been reported to be efficacious in small trials,27,46–49 and it was consistent with the CSP guidelines for the management of WAD. However, the evidence base supporting these treatments is not high quality, and it is possible that there may be other techniques, for example CB therapy, that would be more effective. These require evaluation in future trials.

It is reassuring that the great majority of people we studied eventually recovered from their injury. We found that injury severity had some impact on treatment effectiveness, with the more serious injuries benefiting more from additional treatment. Serious adverse events were rare, and all of the treatments applied had an acceptable safety profile. Although there is some evidence of ceiling effects in the NDI score, we do not believe these to have had a substantial effect on our being able to detect clinically important differences.

There has been a long-standing interest surrounding the possible negative effect compensation has on outcome of whiplash injuries. 21,85–92 We investigated this from two perspectives. The number of people who pursued compensation was high, > 90% of participants who responded in both the Step 1 and Step 2 cohorts. Neither of the advice treatments affected the proportion of people pursing compensation. There was a suggestion in the analysis of Step 2 that physiotherapy treatment appeared more effective in those not pursuing a claim for compensation, but given the very small number of participants in this subgroup, this is uncertain.

Future research questions

Further research is needed to better understand the importance of whiplash injuries in the spectrum of injury that are managed by the NHS. Although not part of the remit of the commissioned trial, the MINT data will be important in determining costs of whiplash injury by different grades of severity, and to determine whether or not poor recovery can be predicted with reasonable accuracy at different time points in the recovery trajectory. A better understanding of the basic epidemiology of disability after whiplash injury will also serve to develop new treatments. We would recommend further research to develop group-based CB programmes, which have been successful in the management of subacute and chronic lower back pain, for the management of WAD. Understanding the optimal time point to administer these programmes will be an important element of future research. Improved educational strategies in the ED, particularly those that can be given to a patient and result in a behavioural change, seem suitable candidates for further testing and development. Improving the content and relevance of The Whiplash Book, for example by addressing problems related to travel anxiety and other psychological manifestations related to stress, may improve effectiveness of the educational materials.

Contribution of authors

Sallie Lamb: chief investigator, study conception and design, development of interventions, supervision, writing and reviewing report.

Mark Williams: study management, development of interventions, writing and reviewing report.

Esther Williamson: development of interventions, study management, writing and reviewing report.

Simon Gates: study supervision, writing and reviewing report.

Emma Withers: study management, writing and reviewing report.

Shahrul Mt-Isa: designing and conducting statistical analysis, writing and reviewing report.

Deborah Ashby: study conception and design, study supervision, designing and conducting statistical analysis, writing and reviewing report.

Emanuela Castelnuovo: designing and conducting economic analysis, writing and reviewing report.

Martin Underwood: study conception and design, study supervision, writing and reviewing report.

Matthew Cooke: study conception and design, study supervision, writing and reviewing report.

Trial team