Spinal manipulation for acute low‐back pain
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
The objective of this review is to examine the effectiveness of SMT on pain, functional status and recovery at the short, intermediate and long‐term follow‐up measurements as compared to the control treatments (e.g. no treatment, sham and all other treatments) for adults with acute low‐back pain.
Background
Low‐back pain is a common and disabling disorder in western society, which represents a great financial burden in the form of direct costs from loss of work and medical expenses and indirect costs (Dagenais 2008). Therefore, adequate treatment of low‐back pain is an important issue for patients, treating clinicians, and healthcare policy makers. One widely used intervention for low‐back pain is spinal manipulative therapy (SMT), which has been examined in numerous randomised controlled trials (RCTs). These trials have been summarized in recent systematic reviews (Bronfort 2004a; Brown 2007; Brox 1999; Cherkin 2003), which have formed the basis for recommendations in clinical guidelines (Chou 2007; Staal 2003; van Tulder 2006; Waddell 2001). However, these recommendations are largely based on an earlier version of this Cochrane review (Assendelft 2003; Assendelft 2004). That review concluded that SMT for acute low‐back pain was superior to sham manipulation and therapies thought to be ineffective or harmful; however, the effect sizes were small and arguably not clinically relevant. Furthermore, these estimates were based mainly on small studies with a high risk of bias. On the other hand, there were no clear differences between SMT and standard therapies, such as general practitioner care, analgesics, exercise, or back schools, for short‐ or long‐term pain relief or functional improvement.
Given the wide use of SMT throughout the world, it is included in many national guidelines for the management of acute low‐back pain (Koes 2001; van Tulder 2004); however, recommendations vary. In most guidelines, SMT is considered to be a therapeutic option in the acute phase of a low‐back pain episode. The USA, UK, New Zealand and Danish guidelines consider SMT a useful treatment for acute low‐back pain. In the Dutch, Australian and Israeli guidelines, SMT is not recommended for the acute phase (van Tulder 2006).
Our goal is to update the previous Cochrane systematic review following the most recent guidelines developed by the Cochrane Collaboration, in general (Higgins 2009) and the Cochrane Back Review Group (Furlan 2009), in particular. In contrast to the previous Cochrane review (Assendelft 2003; Assendelft 2004), the update has been split into two parts by duration of the complaint, namely acute and chronic low‐back pain. The review on chronic low‐back pain (pain lasting for three months or longer) has been completed and is currently under review (Rubinstein 2009). The present review is to report on the effectiveness of SMT for acute low‐back pain.
Description of the condition
Low‐back pain is defined as pain and discomfort, localised below the costal margin and above the inferior gluteal folds, with or without referred leg pain. Acute low‐back pain is defined as the duration of an episode persisting for no longer than six weeks. This condition is considered to be typically self‐limiting, with a recovery rate of 90% within six weeks of the initial episode, while 2% to7% develop chronic low‐back pain (van Tulder 2006). Non‐specific low‐back pain is operationally defined as low‐back pain not attributed to a recognisable, known specific pathology (e.g. infection, tumour or fracture).
Description of the intervention
In this review, we are considering SMT to be any hands‐on treatment that includes manipulation, mobilization or both, of the spine. Mobilizations use low‐grade velocity, small or large amplitude passive movement techniques within the patient's range of motion and control. Manipulation, on the other hand, uses a high velocity impulse or thrust applied to a synovial joint over a short amplitude at or near the end of the passive or physiologic range of motion, which is often accompanied by an audible "crack" (Sandoz 1969). The cracking sound is caused by cavitation of the joint, which is a term used to describe the formation and activity of bubbles within the fluid (Evans 2002; Unsworth 1971). Various practitioners, including chiropractors, manual therapists (physiotherapists trained in manipulative techniques), orthomanual therapists (medical doctors trained in manipulation) or osteopaths use this intervention. However, the focus of the treatment, education, diagnostic procedures used, treatment objectives and techniques, as well as philosophy of the various professions differ. For example, the focus of orthomanual therapy is on correcting abnormal positions of the skeleton and establishing symmetry in the spine through mobilization. Manual therapy focuses on correcting functional disorders of the musculoskeletal system through predominantly passive mobilization and sometimes uses high‐velocity low‐amplitude (HVLA) techniques. Chiropractors, on the other hand, focus on correcting disorders of the neuro‐musculoskeletal system, by using predominantly HVLA manipulative techniques (van de Veen 2005).
How the intervention might work
Many hypotheses exist about the mechanism of action for spinal manipulation and mobilization (Bronfort 2008; Khalsa 2006; Pickar 2002), which to some extent is due to the difference in opinions between the various professional groups. Some have postulated that mobilization and manipulation should be assessed as separate entities given their theoretically different mechanisms of action (Evans 2002). The modes of action might be roughly divided into mechanical and neurophysiological. The mechanistic approach suggests that SMT acts on a manipulable lesion (often called the functional spinal lesion or subluxation), and proposes that forces to reduce internal mechanical stresses result in reduced symptoms (Triano 2001). However, given the non‐nociceptive behaviour of low‐back pain, a purely mechanistic theory alone cannot explain clinical improvement. Much of the literature focuses on the influence on the neurological system, where it is suggested that SMT impacts the primary afferent neurons from paraspinal tissues, the motor control system and pain processing (Pickar 2002). However, the actual mechanism remains debatable (Evans 2002; Khalsa 2006).
Why it is important to do this review
Spinal manipulative therapy is a worldwide, extensively practiced intervention provided by a variety of professions. However, the efficacy of this therapy for acute low‐back pain is not without dispute. We feel that this review, with its comprehensive and rigorous methodology, will provide better insight into this problem. Although numerous systematic reviews have examined the efficacy of SMT for low‐back pain (Airaksinen 2006; Chou 2007), very few have conducted a meta‐analysis, especially for acute low‐back pain. The previous version of the Cochrane review was published in 2004. Since then, numerous RCTs have been identified. It is likely that these newer studies are better designed and conducted, and therefore, likely to provide a less biased assessment. Finally, the methodology for conducting systematic reviews, including the criteria for evaluating the risk of bias, have been substantially updated since publication of the previous Cochrane review, meaning, the results of this update are thought to give a more reliable answer to our research question (Higgins 2009).
Objectives
The objective of this review is to examine the effectiveness of SMT on pain, functional status and recovery at the short, intermediate and long‐term follow‐up measurements as compared to the control treatments (e.g. no treatment, sham and all other treatments) for adults with acute low‐back pain.
Methods
Criteria for considering studies for this review
Types of studies
We will only include truly randomised studies. We will exclude studies using inappropriate randomisation procedures (e.g. alternate allocation, birth dates) as well as studies with less than one day follow‐up.
Types of participants
Inclusion criteria
-
Adult participants (>18 years of age) with low‐back pain with a mean duration (for the study population) for the current episode <6 weeks, meaning the study population had pain for an average of <6 weeks.
-
Patients with or without radiating pain
We will place no limits on the setting (i.e. whether it be from primary, secondary or tertiary care).
Exclusion criteria
Subjects with:
-
Post‐partum low‐back pain or pelvic pain due to pregnancy
-
Pain not related to the low back, e.g. coccydynia
-
Post‐operative studies or subjects with “failed‐back syndrome”
or studies which
-
Examine prevention
-
Are designed to test the immediate post‐intervention effect of a single treatment only, with no additional follow‐up (because we are interested in the effect of SMT beyond one day).
-
Exclusively examine specific pathologies, e.g. sciatica. Of note: Studies of sciatica will be excluded because it is a prognostic factor associated with worse pain, disability, or both (Bronfort 2004; Bouter 1998), especially with SMT (Axen 2005; Malmqvist 2008), and is thought to represent a pathology different than non‐specific low‐back pain.
Types of interventions
Experimental intervention
The experimental intervention to be examined in this review includes both spinal manipulation and mobilization of the spine. Unless otherwise indicated, SMT refers to both modes of "hands‐on" treatments of the spine.
Types of comparison
We will include studies if the study design used suggests that the observed differences were due to the unique contribution of SMT.This excludes studies with a multi‐modal treatment as one of the interventions (e.g. standard physician care + spinal manipulation + exercise therapy) and either a different type of intervention or only one intervention from the multi‐modal therapy as the comparison (e.g. standard physician care alone), since this would make it impossible to decipher the actual effect of SMT.
Comparison therapies will be combined into the following main clusters:
1) SMT versus no treatment, inert interventions or waiting list control
2) SMT versus sham SMT
3) SMT versus all other therapies
4) SMT plus any intervention versus that same intervention alone.
Inert interventions will include, for example, detuned diathermy and detuned ultrasound. The comparison, "all other interventions" will include interventions that are presumed both effective and ineffective for the treatment of acute low‐back pain. Determination of what interventions are considered ineffective (e.g. ultrasound, myofascial therapy, diathermy) and effective (e.g. standard medical care, physiotherapy, exercise) will be based on the literature and our interpretation of those results (Airaksinen 2006; Chou 2007). This will be defined a priori.
Types of outcome measures
Only patient‐reported outcome measures will be evaluated. Physiological measures, such as spinal flexibility or degrees achieved with a straight leg raise test (i.e. Lasegue’s) are not considered clinically‐relevant outcomes and will not be included.
Primary outcomes
-
Pain, measured by a visual analogue or other pain scale (e.g. visual analogue scale (VAS), numerical rating scale (NRS), McGill pain score)
-
Functional status, measured by a back‐pain specific scale (e.g. Roland‐Morris Disability Questionnaire (RMDQ), Oswestry Disability Index (ODI))
-
Global improvement or perceived recovery, measured by an ordinal or dichotomous scale (defined as the number of patients reported to be recovered or nearly recovered).
Secondary outcomes
-
Perceived health status (e.g. the general health perceptions subscale from the SF‐36, the EuroQol thermometer, general health (e.g. as measured on a VAS scale) or similarly validated index)
-
Return‐to‐work
Search methods for identification of studies
Electronic searches
RCTs and systematic reviews will be identified by electronically searching the following databases. The search will be limited to studies published since 2000 because earlier reports were included in the previous Cochrane review.
-
The Cochrane Central Register of Controlled Trials (CENTRAL) from 2000 to present (Appendix 1)
-
MEDLINE from 2000 to present (Appendix 2)
-
EMBASE from 2000 to present (Appendix 3)
-
CINAHL from 2000 to present (Appendix 4)
-
PEDro ‐ current search
-
Index to Chiropractic Literature ‐ current search
The search strategy developed by the Cochrane Back Review Group will be followed using free text words and MeSH headings and will be conducted by a librarian with experience in searching for articles for systematic reviews.
Searching other resources
In addition to the aforementioned, we will also 1) screen the reference lists of all included studies and systematic reviews pertinent to this topic; and 2) search the main electronic sources of ongoing trials (National Research Register, meta‐Register of Controlled Trials; Clinical Trials).
Data collection and analysis
Two review authors (SMR, CBT) will independently perform certain aspects of the review, including selection of studies, data extraction and assessment of risk of bias. A third review author (MWvT) will be contacted if an arbiter is necessary.
Selection of studies
We will screen the titles and abstracts from the search results. Potentially relevant studies will be obtained in full text and independently assessed for inclusion. Disagreements will be resolved through discussion. Only full papers will be evaluated. Abstracts and proceedings from congresses or any other "grey literature" will be excluded. No language restrictions will be imposed.
Data extraction and management
A standardized form will be used to extract data from the included papers. The following will be extracted: study characteristics (e.g. country where the study was conducted, recruitment modality, source of funding, risk of bias), patient characteristics (e.g. number of participants, age, gender), description of the experimental and control interventions, co‐interventions, duration of follow‐up, types of outcomes assessed, and the authors' results and conclusions. Key findings will be summarized in a narrative format. Data relating to the primary outcomes will be assessed for inclusion in the meta‐analysis. Change scores (means and standard deviations) will be extracted if presented by the majority of the studies, otherwise, final value scores will be used. Outcomes will be assessed at one, three, six and twelve months and data examined according to the time closest to these intervals.
Assessment of risk of bias in included studies
The risk of bias assessment for RCTs will be conducted using the twelve criteria recommended by the Cochrane Back Review Group (Furlan 2009). These criteria are standard for evaluating effectiveness of interventions for low‐back pain (Appendix 5; Higgins 2009). The criteria will be scored as "yes", "no" or "unclear" and reported in the Risk of Bias table. In all cases, an attempt will be made to contact authors for clarification of methodological issues if necessary or for unpublished data. A study with a low risk of bias will be defined as fulfilling six or more of the criteria items, which is supported by empirical evidence (van Tulder 2009), and with no fatal flaw, which is defined as those studies with 1) a drop‐out rate greater than 50% at the first and subsequent follow‐up measurements; or 2) clinically‐relevant baseline differences for one or more primary outcomes (i.e. pain, functional status) indicating unsuccessful randomisation. Quantitative data from these studies will be excluded from the meta‐analyses. The reviewers will not be blinded to authors of the individual studies, institution or journal.
While the criteria used in this review are quite standard, they may raise objections by some because blinding the patient and practitioner to treatment allocation is nearly impossible. Furthermore, given that the primary outcomes assessed in this review are all subjective measures (i.e. pain, functional status, perceived recovery), any attempt to blind the outcome assessor is considered irrelevant because the patient is viewed to be the outcome assessor when evaluating subjective measures. Therefore, if the patient is not blinded, the outcome assessor is also considered not blinded. However, to drop these items from the assessment is to negate the observation that “blinding” provides less biased data (defined as performance or detection bias).
Measures of treatment effect
Pain and functional status will be examined as a standardized mean difference (SMD) between the treatment groups. However, if it is determined that pain is measured by a VAS or NRS in all studies, this outcome measure will be examined as a mean difference (MD), because it is thought that this is easier for clinicians to interpret and results for pain from the individual studies will be converted to a 100‐point scale. However, if it is found that most studies use a VAS or NRS, a decision will be made by the research team about how to examine this construct, in order to maximize data use. A negative effect size will indicate that SMT is more beneficial than the comparison therapy (if final value scores are used in the meta‐analyses), meaning subjects have less pain and better functional status. For dichotomous outcomes (i.e. recovery, return‐to‐work), a risk ratio (RR) will be calculated and the event defined as the number of subjects recovered or returned‐to‐work. A positive RR indicates that SMT leads to a greater chance of recovery or return‐to‐work. A fixed‐effect model will be used for all analyses unless a substantial amount of heterogeneity remains unexplained by the sub‐group or sensitivity analyses, in which case the results will be reported using a random‐effects model. Funnel plots will be examined for publication bias. For each treatment comparison, an effect size and a 95% confidence interval (CI) will be calculated. All analyses will be conducted in Review Manager 5.0.
Assessment of clinical relevance
The determination of clinical relevance will be evaluated by one question, "Is the size of the effect clinically relevant?". Levels of clinical relevance,measured by the pooled effect sizes, are defined as: 1) Small: MD <10% of the scale (e.g. <10mm on a 100 mm VAS); SMD <0.4; relative risk, <1.25 or >0.8 (depending on whether it reports risk of benefit for the intervention or control group); 2) Medium: MD 10 to 20% of the scale, SMD from 0.41 to 0.7, relative risk between 1.25 to 2.0 or 0.5 to 0.8; 3) Large: MD >20% of the scale, SMD > 0.7, relative risks >2.0 or <0.5 (Higgins 2009).
Unit of analysis issues
In cases where three or more interventions are evaluated in a single study, a single “pair‐wise” comparison will be calculated where clinically logical. This step is necessary in order to correct for error introduced by “double‐counting” of subjects of “shared” interventions in the meta‐analyses. In cases where two different interventions cannot be fused to form a single pair‐wise comparison and where the interventions are clearly distinct from one another, the number of subjects for SMT (i.e. the shared group) will be halved. In an unrelated issue, data from cross‐over trials will be used prior to the crossover of the intervention only.
Dealing with missing data
In cases where data are reported as a median and interquartile range (IQR), we will assume that the median is equivalent to the mean and the width of the IQR equivalent to 1.35 times the standard deviation (Higgins 2009, section 7.7.3.5). In studies where a range is presented along with the median instead of a IQR, the standard deviation is estimated to be one‐quarter of the range; however, we recognize that this method is not robust and is potentially subject to error (Higgins 2009, section 7.7.3.6). Where data are reported in a graph and not in a table, we will estimate the means and standard deviations. When standard deviations are not reported, we will attempt to contact the author. If the standard deviation for follow‐up measurements is missing, we will use its baseline measure for the subsequent follow‐ups. Finally, if no measure of variation is reported anywhere in the text, we will base the estimate of the standard deviation upon other studies with a similar population and risk of bias.
Assessment of heterogeneity
Heterogeneity will be explored in two manners, by vision (eye‐ball test) and formally tested by the Q‐test (chi‐square) and I²; however, the decision regarding heterogeneity is dependent upon the I² (Higgins 2009) and the cut‐off will be 40%. If the results are thought to be too heterogeneous, such that pooling is thought to be meaningless, then the effect of the interventions will be described.
Assessment of reporting biases
If sufficient studies are available, funnel plots will be constructed for the primary outcomes in order to investigate possible publication bias.
Data synthesis
The overall quality of the evidence and strength of recommendations will be evaluated using GRADE (Guyatt 2008) and discussed by three principal members of the group (SMR, CBT, MWvT). The quality of the evidence for a specific outcome will be based upon the trial's performance against five principal domains: 1) limitations in design (downgraded if >25% of the participants were from studies with a high risk of bias (RoB)), 2) inconsistency of results (downgraded in the presence of significant heterogeneity (I² >40%) or inconsistent findings (in the presence of widely differing estimates of the treatment effect, that is, individual studies favouring both the intervention or control group)), 3) indirectness (i.e. generalizability of the findings; downgraded if >50% of the participants were outside the target group, for example, studies which exclusively examined older subjects or included inexperienced treating physicians), 4) imprecision (downgraded if the total number of participants were less than 400 for each continuous outcome and 300 for dichotomous outcomes) and 5) other (e.g. publication bias). Single studies (N < 400 for continuous outcomes; N < 300 for dichotomous outcomes) will to be considered inconsistent and imprecise and provide “low quality evidence”, which can be further downgraded to "very low quality evidence" if limitations in design or indirectness are also present. Summary of Finding tables will be generated for the primary analyses and for the primary outcome measures only, regardless of statistical heterogeneity, but when present, this will be noted. The quality of the evidence will be defined as:
High quality: Further research is very unlikely to change the estimate of effect or our confidence in it. There are sufficient data with narrow confidence intervals. There are no known or suspected reporting biases.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate; one of the domains is not met.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change it; two of the domains are not met
Very low quality: Great uncertainty about the estimate; three of the domains are not met.
No evidence: No evidence for this outcome from RCTs.
Subgroup analysis and investigation of heterogeneity
Regardless of possible heterogeneity of the included studies, the following stratified analyses will be conducted: 1) By control groups as defined in Types of intervention (see Types of comparisons) and specifically, by passive or ineffective therapies (e.g. diathermy, ultrasound, single counselling session with advice on back pain) versus active or effective therapies (e.g. exercise, standard medical care, physiotherapy); and 2) by time, that is, short‐term (closest to 1 to 3 months), intermediate (closest to 6 months) and long‐term follow‐up (closest to 12 months). Studies reporting outcomes beyond nine months will be considered long‐term.
Sensitivity analysis
The following sensitivity analyses are planned a priori and will be conducted in order to explain possible sources of heterogeneity between studies and to determine the robustness of the original analyses when sufficient studies are available: 1) For risk of bias; 2) for studies with an adequate allocation procedure; 3) by duration of the low‐back pain (i.e. studies which include acute and subacute versus studies of exclusively acute low‐back pain); 4) by type of technique (i.e. high‐velocity low‐amplitude manipulation); 5) by type of manipulator (e.g. chiropractor, manual therapist/physiotherapist); and 6) specific therapies (e.g. exercise, usual care). Summary forest plots will be constructed in STATA with these data in order to give a pictorial overview of these results.
