Abstract
Stratified care for back pain involves targeting treatment to subgroups of patients based on their key characteristics such as prognostic factors, likely response to treatment and underlying mechanisms. It aims to tailor therapeutic decisions in ways that maximise treatment benefit, reduce harm and increase health-care efficiency by offering the right treatment to the right patient at the right time. From being called the ‘Holy Grail’ of back pain research over a decade ago, stratified care is becoming the zeitgeist in research and clinical practice. In this chapter, we introduce and evaluate the quality and underpinning evidence for three examples of stratified care for back pain to highlight their general principles, research design issues and clinical practice implications. We include consideration of their merits for implementation in practice. We conclude with a set of remaining, key research questions.
Introduction
A sobering reflection is that despite a general increase in low back pain research study numbers and quality in recent decades, available treatments tend to produce at best, small-to-moderate mean effects , typically in the short term, with none affecting longer-term prognosis . There are several key explanations, explored in detail elsewhere , but one that has spawned a surge in clinical and research interest in the last decade is that of patient heterogeneity or variability in response to treatment. As randomised trials usually focus on average treatment effects in heterogeneous patient groups, they can fail to reveal the wide range of individual responses to specific treatments, from those who benefit a great deal, to those who benefit little or may even be harmed. Thus, a compelling argument for achieving better treatment results is to match groups of patients with the most appropriate treatment for their profile, referred to as stratified care.
Stratified care involves targeting treatment to patient subgroups based on key characteristics such as their prognostic profile, likely response to specific treatment and suspected underlying causal mechanisms. It ‘fast tracks’ patients to appropriate treatment by supporting therapeutic decision making in order to maximise treatment-related benefit, reduce harm and increase health-care efficiency . Low back pain is an ideal clinical condition within which to research stratified care as it includes a heterogeneous population with clear variation in prognosis and numerous treatment options, some of which are costly and associated with higher risks . In addition, most clinicians believe nonspecific low back pain to include a number of distinct conditions and therefore already use pattern recognition and patient profiling to target their treatment decision making . Stratified care is also particularly suited to low back pain given that the sheer numbers of patients make it unsustainable to offer resource-intensive treatments to all. Subgrouping and targeting care for these patients has been a top international research priority for over 17 years . From being called the ‘Holy Grail’ of back pain research over a decade ago , stratified care is becoming the zeitgeist or dominant school of thought in research and clinical practice.
The idea of stratification is more than 50 years old given that, in 1957, Cronbach wrote “we should design treatments not to fit the average person but to fit groups with particular aptitude patterns…on the assumption that aptitude-treatment interactions exist.” . The term ‘stratified medicine’ has been particularly coined in relation to drug targeting in cancer , but globally many clinical fields are progressively moving towards stratified care (e.g., diabetes and cardiovascular risk ) with the ultimate goal of personalised medicine in the future.
There has been a proliferation of studies testing a wide array of approaches to stratifying low back pain patients for treatment, each with different (but sometimes overlapping) philosophies and methods. There are now almost as many systematic reviews on this topic as original studies and collectively these highlight the limitations in the evidence base to date. These include the lack of plausible rationales for the patient groups or classifications , wide variation in the proportions of patients classified into groups , small sample sizes , concerns about spectrum bias as well as the lack of published protocols, trial registration and long-term follow-up . Overall, the reviews either conclude there is limited evidence to support the clinical application of stratified care or they go further to say that available data do not provide evidence of improved patient outcomes . Interestingly, no stratified care approach has yet attempted to subgroup patients across the whole spectrum of available treatments (physical, pharmacological and surgical) . Of particular importance is the fact that most available studies use designs that cannot differentiate between more general predictors of outcome (prognosis regardless of treatment) and predictors of response to specific treatment (treatment effect modifiers) , as very few use the randomised controlled trial (RCT) design.
Reassuringly, there is a growing body of published guidance focussed on the key stages of research in stratified care, including early stages of development or derivation of the subgrouping method, testing and validating it in both narrow (similar clinical setting and population) and broad validation (broader clinical settings and populations) and assessing its impact on patient outcomes, clinical behaviour, resource use and costs . The proliferation of interest in prognostication to guide treatment decisions, in general, and stratified medicine, in particular, has fuelled a recent initiative, the MRC PROGnosis RESearch Strategy (PROGRESS) Partnership ( www.progress-partnership.org ) and a series of helpful publications (e.g., Refs. ). In this chapter, we consider stratified care as broadly one of three approaches whilst acknowledging there are overlaps between them: those based on patients’ prognosis, those based on underlying causal mechanisms and those based on treatment responsiveness; see Fig. 1 .
We realise, through our own experiences, that research studies providing robust evidence for stratified care in low back pain are challenging to design, fund and conduct . For the purposes of this chapter, we consider one key example of each of the three broad approaches summarised above, based on evidence from the most appropriate research design, specifically, at least one high-quality RCT. Summary details are provided in Table 1 .
| Stratified care approach and specific example | Research setting and intended patient population | Key stratification method | Patient groups | Matched treatments | Degree of prescription or flexibility the approach permits | Strength of the evidence base |
|---|---|---|---|---|---|---|
| Prognosis STarT back approach | Primary care/first contact care Patients with non-specific low back pain of all durations, consulting in UK primary care | Self-report brief screening tool, based on risk of persistent disability | Low risk: patient has a good prognosis Medium risk: patient has a possible poor prognosis without additional treatment support High risk: patient has a probable poor prognosis with significant impact and distress | Low risk: reassurance, medication, self-management and clear explanation (legitimise symptoms and discourage over treatment/investigations) Medium risk: as for low risk plus reactivation support using evidence based conservative treatments (eg. exercise, manual therapy) High risk: as for low risk plus deeper biopsychosocial assessment with combined psychologically-informed treatment and physical reactivation | Guidance for broad patient subgroups to receive different types and intensity of treatments. To be used to inform clinical decision-making. | Derivation: Two primary care prospective cohort studies Validation: One high quality RCT supports approach, powered to compare subgroups, with cost-effectiveness analysis Impact analysis: One high quality implementation study in everyday primary care supports approach, powered to compare subgroups, with cost-effectiveness analysis . Broad external validation of screening tool but no published broad validation RCT of approach (one currently in progress). |
| Treatment responsiveness Manipulation clinical prediction rule |
| Based on meeting 4 of 5 CPR criteria:
| Positive on CPR Or Negative on CPR |
|
| Derivation: One cohort study [38] Narrow validation: One high quality RCT supports approach Broad validation: One high quality RCT did not support approach in broader setting using primarily mobilisation rather than manipulation Impact analysis: No studies in chronic low back pain. No cost-effectiveness or implementation studies |
| Underlying mechanisms Classification based Cognitive Functional Therapy | Primary and secondary care Patients with recurrent, persistent non-specific low back pain provoked by postures, movements and activities | A mix of clinical assessment and self-report questionnaires to identify symptom-provoking and modifiable cognitions, movement and lifestyle behaviours | Different levels are integrated as indicated
|
|
| Derivation: Clinical observation, laboratory studies on movement control and cohort studies on predictors of outcome Validation: Studies on inter-tester reliability and laboratory studies of movement patterns support the approach. One high quality RCT supports the approach Impact analysis: No studies in acute low back pain. No cost-effectiveness or implementation studies |
Key examples
Stratification based on prognostic risk
One high-profile and relatively recent approach to stratification is based on a multi-domain prognostic model, where each patient’s risk for developing persistent disabling back pain is determined and used to match patients to treatment. The STarT Back (Subgroups for Targeted Treatment) approach, developed and tested in the UK , allocates patients to one of three subgroups (patients at low, medium and high risk of persistent back pain) and has a growing body of research evidence suggesting clinical benefits when patients at medium and high risk are ‘fast tracked’ to an appropriate course of treatment whilst those at low risk, who have a good prognosis, are steered away from over-investigation and -treatment. The approach is particularly designed to support primary/first-contact care decision making.
Key advantages of the STarT Back stratification approach are its inclusivity (all patients are subgrouped) and simplicity (patients complete a brief nine-item, self-report tool) . The tool uses key prognostic factors in back pain that are considered to be modifiable through treatment, including physical (leg pain, co-morbid pain and disability with walking and dressing) and psychosocial factors (fear, anxiety, pain catastrophising, mood and overall impact). 4
4 Following initial publication (Hill et al., 2008), tool items were reordered to permit simpler scoring (Hill et al., 2011 online appendix). Both self-report and interview formats are available (see www.keele.ac.uk/sbst/onlinetool/ ).
It does not therefore include known, but nevertheless, non-modifiable factors such as back pain episode duration. There is, however, overlap between this stratification model and others. This is because although ‘prognosis’ determines STarT Back tool cut-points to separate patients at low risk from those at medium risk and high risk, patient prognosis is not used to separate medium- from high-risk subgroups. The tool developers used expert consensus to agree on a level of patient distress at which patients were considered likely to specifically require a psychologically informed physiotherapy approach to separate medium risk from high risk (this was agreed on as 4 out of 5 on the psychological subscale of the STarT Back tool). This stratification method therefore is not purely prognostic but also uses, to some extent, expected patient response to treatment.Research has confirmed the tool’s sensitivity to change and discriminant, concurrent, criterion and predictive validity in external samples and in different settings and languages . Given that the prognostic stratification is based on modifiable factors, there is a decrease in the tool’s predictive ability in treatment contexts where those risk factors are effectively targeted , when the tool is used in different settings (e.g., secondary care ) or when it is used for different purposes (e.g., monitoring patient progress ).
The STarT Back approach emerged from a lengthy research programme at Keele University in the UK. The team initially identified important prognostic factors for persistent low back pain and disability using large epidemiological cohort studies of primary care consulters to develop and validate the STarT Back tool . Matched treatments for each patient subgroup were underpinned by available research evidence and expert consensus based on an understanding of each subgroup’s characteristics . For patients at low risk, matched treatment comprises a package of care involving assessment, reassurance 5
5 The Keele research team has collaborated with AXA PPP to produce a patient advice digital video disc (DVD); see www.keele.ac.uk/sbst/onlinetool/ .
, medication advice, self-management advice and a clear explanation to legitimise symptoms and discourage over-treatment or investigation. For patients at medium risk, matched treatment provides re-activation support using evidence-based, conservative treatments offered by physiotherapists (including manual therapy and exercise). For the most complex patients in primary care, those at high risk of persistent disability, matched treatment is psychologically informed physiotherapy , combining physical and psychological treatment approaches. The clinical and cost-effectiveness of the STarT Back stratified care model (both screening and matched treatments) was tested in a large randomised trial (the STarT Back trial, a Phase III trial, n = 850) . In addition, an impact analysis study of before/after design investigated whether the STarT Back stratified care approach is implementable in everyday clinical practice with family doctors, physiotherapists and patients (the IMPaCT Back study, a Phase IV study, n = 922) and whether it brings benefits for patients, clinicians, healthcare processes and costs. Qualitative interviews explored clinicians’ perceptions of using this model in practice .Results of the STarT Back trial showed significant patient benefits and cost-effectiveness primarily due to reductions in work absence . The IMPaCT Back study confirmed that this approach is feasible for use in primary care, changes treatment decision making, improves patient outcomes and is cost-effective . Qualitative interviews revealed that clinicians’ perceptions vary, and in general, physiotherapists are more willing to adopt this new approach than their family doctor colleagues . Early audits of services implementing the STarT Back model suggest that an initial increase in physiotherapy referrals declines over time and that use of the tool means therapists discharge low-risk patients earlier and spend more time with complex patients. One barrier to sustainable implementation is the training needed to deliver the matched treatments and a ‘train the trainer model’ is likely needed to cascade the knowledge and skills in the matched treatment, particularly for patients at high risk. A further limitation is that the designs of the STarT Back and IMPaCT Back studies cannot tease apart whether the subgrouping or matched treatments (or both) led to the beneficial effects observed. Validation studies in other health-care settings are needed. An RCT of the STarT Back approach in a US health organisation, Group Health, is currently in progress, led by Professor Dan Cherkin 6
6 http://www.grouphealthresearch.org/faculty/profiles/cherkin.aspx .
with results expected from 2015.In summary, prognostic stratification based on a multi-domain prognostic model has shown promising results and is being recommended in national and international clinical guidance . However, broad validation of matching patients to treatments based on risk subgroups is needed.
Stratification based on mechanisms
Many mechanism-based classification systems have been advocated in order to deal with the heterogeneity of low back pain and enhance treatment matching. However, few have been tested in randomised trials and most are criticised for being unidimensional (thus failing to reflect the biopsychosocial nature of back pain) and having poor validity . In response to these limitations, a multidimensional classification system has been developed and tested , Classification Based-Cognitive Functional Therapy (CB-CFT), that integrates contemporary evidence about the associations between low back pain and factors within pathoanatomical, neurophysiological, psychosocial, physical and lifestyle domains . Within the different domains, stratification according to CB-CFT is designed to provide direction for improved clinical management.
CB-CFT incorporates a biopsychosocial clinical assessment (interview and physical examination) combined with psychosocial screening and review of radiological and medical investigations where appropriate, in order to triage patients, identify underlying modifiable mechanisms and stratify patients across the five broad domains in order to target care better. There is evidence to support each of the five domains within this approach. For example, subgrouping patients based on ‘mechanical’ or ‘non-mechanical’ pain characteristics suggestive of different underlying neurophysiological pain mechanisms has been shown to have discriminative validity . The use of validated psychosocial screening questionnaires in conjunction with the clinical examination identifies patients with a high psychological risk profile while highlighting individual risk factors in order to target care . A number of laboratory-based studies have confirmed clinical reports that at least two groups of patients, with nonspecific low back pain and ‘mechanical’ pain characteristics, present with different functional postures, movement patterns and motor responses that differentiate them from each other and those without low back pain .
CB-CFT was specifically developed as a stratification approach for targeting treatment in patients with nonspecific low back pain provoked by activities, movements and postures . Modifiable beliefs (such as negative back pain beliefs, fear of movement and pain-related anxiety) and behaviours (such as pain-provoking postures and movement patterns and avoidance and pain behaviours) considered to perpetuate pain and disability, are identified and become the targets for treatment . The primary aims of CB-CFT are to provide an individualised, biopsychosocial understanding of pain, enhance pain-coping strategies through cognitive restructuring and targeted movement training based on movement classification and promote pain self-efficacy and confidence by normalising movements and activities that were previously avoided or reported as provocative. In situations where pain features or psychological co-morbidities are barriers to behavioural change, CB-CFT can be integrated with other medical and/or psychological management.
Underpinning research has assessed the inter-tester reliability of different aspects of CB-CFT and shown substantial agreement between trained clinicians (physiotherapists and medical practitioners) in the identification of patient subgroups based on pain-related functional postures and movement patterns . The findings show moderate-to-substantial levels of agreement in patient subgrouping across the first four domains of the system. However, further research is needed to establish the clinical utility and reliability of the patient subgrouping for all low back pain patients across all domains of the system.
The clinical effectiveness of this stratification approach has been compared with physiotherapy-led exercise and manual therapy in a randomised trial in Norway, in a primary care setting, with 121 nonspecific low back pain patients whose pain was of more than 3 months’ duration, provoked by movements, activities and postures and who had moderate disability levels . At 12 months’ follow-up, significant differences were observed in favour of CB-CFT for disability and pain, as well as other outcomes (fear of movement, mood, work and satisfaction levels) . Whilst no formal cost analysis was conducted, potential cost savings from CB-CFT were suggested by reductions in ongoing health care and sick leave over 12 months. The results of this trial are encouraging, but there are no published data yet available about the implementation of CB-CFT in primary care or secondary care clinical practice nor in case of patients with higher disability levels.
Whilst the broad multidimensional classification system lends itself to any clinician managing patients with low back pain, the CB-CFT stratification approach is arguably best suited to physiotherapy practice, as it requires specific skills across a number of domains . The approach requires clinicians to develop skills to effectively communicate, teach body relaxation strategies, normalise functional movement patterns and discourage pain behaviours, while utilising mindfulness and motivational principles. Previous reports suggest that an average of 100 h of training is required for experienced physiotherapists to reliably implement this approach to patient stratification , and publications, clinical workshops and open access Web-based resources ( www.pain-ed.com ) support this.
In summary, while there is evidence to support both the validity and reliability of patient subgroups based on underlying mechanisms and some evidence for this type of stratified care from RCT (e.g., Ref. ), there are no impact analyses or cost-effectiveness data available. Further high-quality randomised trials, ideally with nested assessments of potential treatment mechanisms, are needed that compare mechanisms-based stratified care with other approaches and in different clinical settings.
Stratification based on treatment respondents
A further approach to stratified care for back pain is to match patients to treatment based on their likely responsiveness to that treatment. Whilst there is some overlap with the approaches previously described, the key to this approach is that it starts with an existing treatment (e.g., manipulation) and matches patients to that treatment, rather than starting with patient characteristics (prognosis or mechanisms) which then determine the best treatment. Existing treatments tend to have small effects across the heterogeneous group of people with nonspecific low back pain . The premise is that matching subgroups of patients to these treatments, rather than providing the same treatment to all, increases treatment effectiveness for relevant patient subgroups. The challenge is to provide evidence of patient features that consistently identify those who respond to a specific treatment.
Research investigating treatment respondents has typically been associated with clinical prediction rules (CPRs) . These rules enable the identification of clusters of patient characteristics associated with prognosis, diagnosis or response to treatment . While clusters of characteristics may be important for identifying treatment respondents, it is important to note that a single feature (e.g., age) could also achieve this and CPRs are not necessarily required .
An example of stratification by treatment response is the development and testing of a CPR to identify treatment respondents to manipulation. Previous trials and systematic reviews have shown that manipulation/mobilisation has, at best, small overall benefits . Flynn and colleagues set out to develop a CPR to identify patients who respond to a specific manipulation (a high-velocity thrust). They collected baseline data on a large number of factors from the history, the physical examination and self-report questionnaires in 71 low back pain patients referred to physical therapy. All patients received a standardised single manipulation on three treatment occasions. Patients in whom back-related disability improved by more than 50% were classified as having a successful outcome. Baseline characteristics associated with successful outcome were duration of symptoms of <16 days, Fear Avoidance Beliefs Questionnaire work subscale score of <19, at least one hip with >35° of internal rotation, hypomobility (stiffness) in the lumbar spine with spring testing and no symptoms distal to the knee. The design was a single-arm study and thus it is not clear whether the CPR identifies patients with a good prognosis regardless of treatment or specifically those who respond well to manipulation. CPRs to identify treatment respondents ideally need to be derived in a randomised trial and analysed using interaction terms; however, these studies are expensive, require large sample sizes and should investigate a limited number of subgroups . Thus, the study by Flynn and colleagues is likely best described as ‘hypothesis generating’ .
A well-conducted randomised trial by Childs and colleagues in the US tested the validity of the CPR developed by Flynn. The authors randomised 131 patients to receive manipulation plus exercise or exercise alone (control). The results showed that patients who were positive on the CPR (i.e., they met four or all of the five CPR criteria) had substantially better response to manipulation. A broad validation study of the manipulation CPR was performed by Hancock and colleagues in Australia . They randomised 240 participants to either manipulative therapy (mobilisation or manipulation) or placebo. This trial failed to find a better response to manipulative therapy in those patients positive on the CPR. There are several explanations, including the use of mobilisation (predominant) and manipulation at the discretion of the clinician in the Australian trial, rather than the single manipulation used in the US studies. Alternatively, the different findings could be due to differences in patients, settings or co-interventions. This trial highlights the challenges in demonstrating broad validation, and thus wider utility, of stratification approaches.
There are several important issues pertinent to CPRs. These include the proportion of patients assessed who meet the CPR criteria (e.g., in Childs et al. , whilst 36% of included patients met the CPR this was only 13% of all patients screened) and lack of guidance to clinicians in managing patients who do not meet the CPR criteria. The latter problem can be, at least in part, overcome through the use of treatment-based classification algorithms that match patients to one of several treatments (e.g., Refs. ). There is a real need for impact analysis studies that compare CPR-based stratification with usual care rather than only testing interaction in RCTs evaluating the specific intervention, as well as external validation in other settings, populations or spectrum of low back pain patients.
In summary, stratification of care based on identifying groups of low back pain patients who respond to specific treatments is still at an early stage of research development. Very few CPRs have been tested in RCTs and even the most promising approaches lack evidence from broad validation studies. An important limitation to implementing stratification approaches based on response to individual treatments (e.g., manipulation) is that typically only a small proportion of patients will meet the CPR criteria. More inclusive approaches based on treatment responsiveness that can guide decision making for the majority of low back pain patients may be easier to implement in clinical practice.
Related posts:
Trajectories of low back pain
Psychological factors and treatment opportunities in low back pain
Effects of workplace, family and cultural influences on low back pain: What opportunities exist to address social factors in general consultations?
Is it all about a pain in the back?
Why and how back pain interventions work: What can we do to find out?
Low back pain research – Future directions
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
