Objectives
This is a protocol for a Cochrane Review (intervention). The objectives are as follows:
The objectives of this systematic review, conducted using a collaborative review model, are to:
Assess the effectiveness of exercise treatment (overall) in adults with chronic non‐specific low back pain on important individual health outcomes: pain, functional limitations, health‐related quality of life, depression, and adverse effects versus comparison treatments: (a) placebo, sham, or attention control, (b) no trial treatment (including waiting lists, control groups described as having no treatment provided, usual/normal care not controlled by the trial available to all treatment groups, or when the exercise and comparison groups receive the same co‐interventions, allowing the effect of exercise treatment to be isolated), and (c) other conservative treatments (eight categories).
Estimate the treatment effects and associated uncertainty for comparisons of different specific types of exercise treatment in adults with chronic non‐specific low back pain to each other, and to each comparison treatment, using direct and indirect evidence with network meta‐analysis.
Estimate the treatment effects and associated uncertainty for comparisons of treatments composed of different exercise type categories, design, delivery, dose, and additional treatment components, and their combinations, using direct and indirect evidence with component network meta‐analysis.
Background
There is little debate about the global impact of low back pain on people and healthcare systems; this condition has remained one of the leading causes of disability world‐wide for more than 30 years (Chen 2022; GBD 2017). Low back pain results in enormous direct healthcare and lost productivity costs (Dagenais 2008; GBD 2017; Hayden 2009; Lim 2012; Martin 2008; Wu 2020), and arguably, even greater indirect societal costs (Alonso‐García 2020; Grabovac 2019). The majority of the social and economic costs associated with low back pain are attributable to people who experience prolonged disability through the development of chronic (persistent or recurrent) low back pain (Hartvigsen 2018).
Despite a tremendous volume of research in the area, there remains uncertainty about the most effective treatment approaches. Recent work suggests that management of low back pain with exercise is associated with lower healthcare system costs and improvements in quality adjusted life years when compared to usual care for acute and chronic patients (Miyamoto 2019). Our previous Cochrane Review found that exercise was likely to be effective for treating chronic low back pain (Hayden 2021a). Our companion network meta‐analysis compared different types of exercise treatments to each other (Hayden 2021b). Hayden 2021b found that all investigated exercise types were more effective than minimal treatment, and that Pilates, McKenzie exercises, and functional restoration were more effective than other types of exercise treatment for reducing pain intensity and functional impairment.
Description of the condition
Low back pain is defined as pain, muscle tension, or stiffness localised below the costal margin and above the inferior gluteal folds, with or without pain referred to one or both legs (Dionne 2008). Most individuals who experience low back pain are labelled as having non‐specific low back pain, which is defined as symptoms not attributable to a recognisable, known, specific pathology (for example, fracture, disc herniation, spondylolisthesis, spondyloarthritis, infection, neoplasm, or metastasis). Several different structures of the back have been implicated in symptoms of non‐specific low back pain, including the musculature, joints, and discs; but also psychosocial factors, including maladaptive pain coping behaviours, high baseline functional impairment, presence of psychiatric comorbidities, and low general health status (Chou 2010). These psychosocial factors have been associated with persistent, disabling low back pain.
In this review, we consider the effectiveness of exercise treatment for individuals whose symptoms persist beyond the acute stage. Previously, research studies and practice guidelines defined individuals with low back pain episodes persisting beyond the acute stage as subacute and chronic. In this review, we focus on chronic low back pain, which we have defined as pain, muscle tension, or stiffness lasting longer than 12 weeks (chronic persistent) or recurrent low back pain defined as two episodes in a year, lasting more than 24 hours, with more than 30 days pain‐free between (chronic recurrent).
Individuals who experience a low back pain episode often continue to have long‐standing pain. A systematic review of cohort studies found that 33% of people with low back pain recover in the first three months, but 65% still report pain after one year (Itz 2013). Many factors, including biophysical, psychological, social and genetic factors, and comorbidities can contribute to disabling, chronic low back pain (Hartvigsen 2018). Factors associated with delayed recovery of chronic low back pain include previous sick leave due to low back pain, high disability or pain levels at the onset of chronicity, low levels of education, perceived risk of persistent pain, and psychological factors, such as depression, anxiety, catastrophising, and self‐efficacy (Costa 2009; Hartvigsen 2018). Specific predictors of poor outcome in people with chronic low back pain have been inconsistent across prognosis studies. Chronic low back pain appears to be the result of nociceptive input combined with multiple complex factors (Hayden 2009; Heitz 2009).
Description of the intervention
Exercise treatment is widely recommended for the management of low back pain. Exercise treatments include structured physical activity, usually prescribed or planned by a health professional, which include conducting specific activities, postures, or movements (or a combination (Caspersen 1985)). Exercise treatments can be classified as cardiorespiratory/aerobic (e.g. walking, swimming), resistance (e.g. general strength training, core strengthening), flexibility (e.g. static or dynamic stretching), neuromotor training (e.g. balance or coordination training), or combination/multimodal, which involves more than one class. Examples of specific exercise treatment types include McKenzie therapy, Tai Chi, rock climbing, yoga, or horse simulator training.
Exercise treatments are complex interventions, characterised by several interacting components. They are heterogeneous in treatment design (e.g. standard, individualised), dose (duration, frequency, intensity), delivery format (e.g. clinician supervised in person or remote, one‐on‐one, or group), and may be combined with other conservative treatments (e.g. manual therapy, pain medications), where exercise is a major treatment component. We will include trials that compare different doses or delivery approaches of the same exercise treatment (e.g. twice‐weekly yoga versus three times weekly yoga).
How the intervention might work
Various specific types and components of exercise treatment are expected to be associated with different hypothesised mechanisms of effect (Wun 2021). Exercise treatment may provide benefits to people with chronic low back pain through mechanical, psychosocial/cognitive, neuro‐physiologic, and cardiometabolic mechanisms. Potential mechanical mechanisms are discussed most frequently, such as increasing muscle strength or endurance, flexibility, range of motion, or motor control. These mechanisms act through the voluntary contraction of specific muscle groups, movement of the whole body, activities that improve postural musculature, stabilisation, or coordination (or a combination (Powell 2011)). Proposed psychosocial mechanisms for exercise treatment of chronic low back pain include decreasing fear avoidance behaviours, kinesiophobia, or competing pain behaviours, and by improving pain self‐efficacy, mood, or coping strategies, which may lead to decreased pain and improved function. Neurophysiologic mechanisms hypothesised for exercise treatments include increased blood flow/circulation, decreasing inflammation, or increased endorphins. Cardiometabolic mechanisms of the effect of exercise treatments have also been proposed to improve general health and fitness.
It has been proposed that exercise treatments act through one or more of these proposed mechanisms to accelerate recovery, reduce pain, improve function, and assist with return to usual activities. However, many studies fail to clearly describe the detailed intervention and assumed mechanisms. Researchers have called on the back pain trialist community to prespecify their treatment target to improve the match between intervention and desired outcome (Wood 2020; Wood 2021).
Why it is important to do this review
Exercise is a common approach to the treatment of chronic low back pain. Clinical practice guidelines, including the American College of Physicians (ACP (Qaseem 2017)) and the UK National Institute for Health and Care Excellence (NICE (National Guideline Centre 2016)), recommend exercise as the first line of care for chronic low back pain, and recommend exercise programmes that take individual needs, preferences, and capabilities into account. However, the most effective components of exercise interventions for people with chronic low back pain have yet to be identified. In our 2021 update of this Cochrane Review investigating the effectiveness of exercise for treating chronic low back pain, we included 249 randomised trials (24,486 participants), with 203 trials providing sufficient data for meta‐analyses (19,633 participants (Hayden 2021a; Hayden 2021b), which followed earlier versions of this review (Hayden 2005a; Hayden 2005b; Hayden 2012)). In the 2021 review update, 142 randomised trials compared exercise to non‐exercise comparisons (117 with data for meta‐analyses), and 151 trials included comparisons of different types of exercise treatments. Despite considering study‐level sources of clinical heterogeneity, the studies in the meta‐analyses were heterogeneous. However, we found consistent results across primary, subgroup, and sensitivity analyses to draw conclusions about the effectiveness of exercise for the treatment of chronic low back pain. There was moderate‐certainty evidence that exercise treatment is probably more effective than no treatment, usual care, or placebo for pain intensity and functional limitation outcomes. For pain intensity outcomes, our findings were most compatible with a clinically important difference of 15 points out of 100 compared to no treatment; for functional limitation outcomes, the mean effect did not meet our prespecified threshold for minimal clinically important difference.
There is low‐ to moderate‐certainty evidence that there is a small difference in pain and functional limitations for exercise treatment compared with other conservative treatments. Data from 64 trials (6295 participants) comparing exercise treatment to other conservative treatments found exercise to be more effective, although the effect size was small and not clinically important overall (9 points improvement in pain; 4 points improvement in functional limitations out of 100); comparisons with some other specific conservative treatments (e.g. electrotherapy, education alone) showed exercise treatments to have larger improvements, compatible with a clinically important difference.
Adverse effects of exercise treatment were rarely reported. When they were reported, those likely to be related to the intervention mainly included increased low back pain and muscle soreness.
Many new trials have been published since the 2021 review that investigated the effectiveness of exercise treatment for chronic low back pain. While it is difficult to keep up with new evidence in this broad field, keeping the evidence together in one review allows insights and analyses that would not be possible if we only undertook focused exercise‐type specific reviews, even though this would be more feasible. Furthermore, an inclusive approach to evidence synthesis results in less duplication of effort, research waste, and can also reduce confusion for decision‐makers by avoiding conflicting conclusions from smaller, focused systematic reviews that use heterogeneous methods.
We are publishing a new protocol for this review update due to substantial changes in our proposed methods. We will conduct an optimised updated literature search, extract new outcomes, and conduct additional planned analyses. We will conduct the review update using a collaborative review model. In our proposed analyses, including pairwise meta‐analysis, and standard and component network meta‐analysis, we will use direct and indirect evidence to compare the effectiveness of specific exercise program types, and assess the importance of component exercise treatment characteristics individually, and in combination. Our overarching Cochrane Review will be complemented by focused Cochrane Reviews, conducted by teams collaborating to ensure similar robust methods, quality control measures, and opportunities for data sharing.
This review update is being conducted with editorial guidance from the Cochrane Musculoskeletal Group, and will be conducted according to the guidelines recommended by the Cochrane Musculoskeletal Editorial Board (Ghogomu 2014).
What is different about the collaborative review model
The collaborative review model (described in Hayden 2022) aims to increase collaboration within the evidence synthesis community in a given research field while following robust, standard systematic review methods. Many teams work together to complete an overarching review (i.e. Exercise therapy for chronic low back pain), and respective sub‐reviews (e.g. yoga for chronic low back pain; aerobic exercise for chronic low back pain), using co‐ordinated review processes, while sharing resources, tasks, and tools to limit duplication of effort and increase efficiency.
The collaborative review model will include conducting the overarching review of all evidence evaluating the effectiveness of exercise treatments for chronic low back pain, with nine overlapping sub‐reviews of specific exercise treatment types, or a focused chronic low back pain population, and an evidence update for a clinical practice guideline. Conducting the review as part of this collaborative review model will provide important individual and complementary evidence. We will use the same comprehensive methods and criteria for all the overarching and focused reviews. This will allow comprehensive comparison of the effectiveness of various exercise treatment approaches and types with network meta‐analysis, as well as facilitate conduct of the sub‐reviews, training the next generation of reviewers, and building research capacity and community in the research field.
Objectives
The objectives of this systematic review, conducted using a collaborative review model, are to:
Assess the effectiveness of exercise treatment (overall) in adults with chronic non‐specific low back pain on important individual health outcomes: pain, functional limitations, health‐related quality of life, depression, and adverse effects versus comparison treatments: (a) placebo, sham, or attention control, (b) no trial treatment (including waiting lists, control groups described as having no treatment provided, usual/normal care not controlled by the trial available to all treatment groups, or when the exercise and comparison groups receive the same co‐interventions, allowing the effect of exercise treatment to be isolated), and (c) other conservative treatments (eight categories).
Estimate the treatment effects and associated uncertainty for comparisons of different specific types of exercise treatment in adults with chronic non‐specific low back pain to each other, and to each comparison treatment, using direct and indirect evidence with network meta‐analysis.
Estimate the treatment effects and associated uncertainty for comparisons of treatments composed of different exercise type categories, design, delivery, dose, and additional treatment components, and their combinations, using direct and indirect evidence with component network meta‐analysis.
Methods
Criteria for considering studies for this review
Types of studies
We will include published reports of completed randomised trials, with no language or date restrictions. We will include only randomised trials, as they are the gold standard to assess the effectiveness of health‐related interventions.
Types of participants
We will include studies involving adult participants (minimum age 16 years, with the majority of participants 18 years and older) with chronic non‐specific low back pain of 12 weeks or more duration (defined as mean back pain duration of the study group of at least 12 weeks).
We will exclude studies that involved individuals with low back pain likely to have been caused by specific pathologies (including radiologically confirmed disc herniation, spinal stenosis, piriformis syndrome, fracture, spondyloarthritis, infection, neoplasm, or metastasis). We will not consider osteoarthritis, osteoporosis, or disc bulge as specific pathologies for this review, studies of pregnant or post‐surgical populations, and studies that focus exclusively on acute exacerbations of chronic low back pain (i.e. flares (Costa 2019)).
Studies of non‐specific low back pain often include participants with a mixture of individual and symptom characteristics, all commonly managed using a variety of conservative treatments. We will only include studies of participants with leg pain if non‐specific back pain is the main complaint for the majority of participants.
We will include studies of participants who were recruited from any setting, including healthcare, occupational, general, and mixed populations. We will characterise the population sources as healthcare (primary, secondary, or tertiary care centres), occupational (participants presenting to occupational healthcare facilities or personnel in compensatory situations), or from a general or mixed population (e.g. including individuals recruited by newspaper advertisements) to differentiate the studies of participants in typical treatment settings (healthcare and occupational) from those including individuals with low back pain who may not normally present for treatment.
This broad definition of adult non‐specific low back pain reflects the definition used in most trials in the field, so is likely to result in sufficient clinical comparability across all direct comparisons included in our review. We will conduct subgroup analyses related to population sources/settings and characteristics for the effect of exercise treatment overall (population settings, duration of chronic low back pain), and test this transitivity assumption in our network meta‐analyses (see Subgroup analysis and investigation of heterogeneity). In secondary analyses, we will separately consider the effectiveness of exercise treatment for study populations that included any study participants with leg pain or neurological symptoms, and populations with moderate‐term (12 weeks to three years) and long‐term (longer than three years) symptom durations of chronic low back pain at baseline (Dunn 2006).
Types of interventions
Exercise treatments include structured physical activity (bodily movement produced by skeletal muscles resulting in energy expenditure) usually prescribed or planned by a health professional that include specific activities, postures, or movements (or all (Caspersen 1985)). This encompasses a heterogeneous set of treatments with a goal to reduce pain and improve functioning. We will include studies of exercise treatments with a duration of at least one week.
We will classify and analyse the exercise treatments in two distinct ways: (1) according to the specific (named) type, typically described by the trial researchers, and (2) according to the component parts of the exercise program. Eligible specific exercise treatments will include aerobic, core strengthening, general strength training, McKenzie therapy, motor control exercise, Pilates, Qigong, stretching or flexibility/mobilising exercises, Tai Chi, yoga, mixed exercise (three or more types), and other (Table 1). We will include exercise treatments that include a combination of specific exercise types. First, we will document all specific exercise types included in each exercise treatment group, regardless of the proportion of the exercise treatment they comprise. Then, we will assign the dominant exercise type for each treatment group, or the group will be designated as mixed (when exercise treatments incorporated two or more exercise types, and one did not clearly predominate).
1. Exercise types, classification description, and categories for analysis.
| Specific exercise type | Classification description or guidance | Category for analysis (ACSM*) |
| Aerobic | Aerobic exercises are exercises involving major muscle groups designed to improve cardiorespiratory fitness; examples include walking, cycling, or swimming. | Cardiorespiratory |
| Core strengthening | Core exercises are back‐specific resistance exercises that aim to strengthen the muscles of the spine and pelvis; these exercises include balance, coordination, core strengthening, stabilisation exercises, or sling exercises, but exclude whole‐body (generalised) stability, or strengthening exercises. | Resistance |
| General strength training | General strength training exercises use load bearing or resistance exercises, possibly including weights, weight machines, elastic cords, or body weight, to increase an individual's ability to exert or resist force; designed to strengthen the whole body (can include, but not targeted exclusively at the low back). | Resistance |
| McKenzie therapy | McKenzie therapy or method is a program according to the specific McKenzie approach, including sustained positions or repeated movements based on clinical observations, such as changes in pain location or movement restriction (directional preference); it also includes postural training, education, and self‐management. | Flexibility; combination |
| Motor control exercise | Motor control exercises use the principles of motor learning (e.g. feedback, segmentation, simplification) to retrain control and co‐ordination of the trunk muscles; this may include motor skill training of the deep muscles of the spine, specifically transversus abdominus, multifidus, and the pelvic floor muscles, restoration of co‐ordination of the deep and superficial muscles during static and dynamic tasks, co‐ordination of breathing and continence with trunk control strategies, progressing to function. | Neuromotor |
| Pilates | Pilates is based on the Pilates principles of centring, concentration, control, precision, flow, breathing, and posture. It is a system of repetitive exercises performed on a mat or other Pilates‐specific equipment to promote strength, flexibility, posture, and mental awareness. | Resistance; combination |
| Qigong | Qigong is a Chinese mind‐body practice of exercise including body‐posture and movement (slow flowing movements), breathing exercises, and meditation. | Neuromotor; combination |
| Stretching or flexibility/mobilising exercises | Active exercises using positions held for varying periods of time (static) or performed with movement (dynamic) to improve range of motion, using aids or not (not passive mobilisations or stretches performed by a therapist). Depending on the location and type, they can be focused on back‐specific structure or whole body. | Flexibility |
| Tai Chi | Tai Chi is a whole‐body Chinese martial art that may include sequences of very slow controlled movements and breathing, and focused attention. | Neuromotor; combination |
| Yoga | Yoga is a practice with many branches and styles, which may involve physical poses (asanas), controlled breathing (pranayama), and meditation (dhyana). The overall intention may be to create union between body, mind, and spirit. | Flexibility; neuromotor; combination |
| Mixed exercise (three or more types) | Mixed exercises are those that incorporate three or more types of exercises, and in which one does not clearly predominate or form the majority of the exercise treatment. Mixed exercises can be whole body, low‐back specific, or both, depending on the types and combination of exercises involved. | Combination |
| Other | Exercise type not described above. Provide explanation. | Unknown |
| Unknown exercise type | No details about exercise type are provided. | Unknown |
*ACSM = American College of Sports Medicine categories of exercise and physical activity; suggested primary classification and for sensitivity analysis
Second, we will further classify exercise treatments according to the component parts of the exercise programme (deconstructing the treatment for description and component network meta‐analysis). This will include classifying the exercise treatment according to the American College of Sports Medicine exercise categories: cardiorespiratory/aerobic, resistance, flexibility, neuromotor training, or combination/multimodal (Table 1). We will also classify exercise treatments by programme design (individually or standard designed), delivery type (independent, independent with follow‐up, group‐supervised, group‐supervised with follow‐up, individually‐supervised), dose (low dose, high dose; considering length, number of sessions, programme duration, and adherence), and inclusion of additional treatments. We will define the design of the exercise treatment as individually‐designed, when the treating therapist completed a clinical history and physical examination, and delivered an exercise treatment specifically designed for the individual participant, or standard design, when a fixed exercise treatment was delivered to all participants (may vary only in intensity or duration). We will categorise the delivery type as: (1) independent, when the participant met with the treating therapist once, and then participated in the treatment at home without supervision; (2) independent with follow‐up, when the participant met with the treating therapist initially, participated in the treatment at home, then received follow‐up with the therapist; (3) group‐supervised, if the participant attended supervised group therapy sessions with two or more participants; (4) group‐supervised with follow‐up, if the participant attended supervised group therapy sessions, then received follow‐up with the therapist; and (5) individual, when the participant attended one‐on‐one sessions with the treating therapist. We will describe the type of therapist, and whether a graded activity approach was used in the exercise treatment. Finally, we will indicate all non‐exercise co‐interventions that were included in the exercise treatment (Table 2).
2. Exercise design and delivery characteristics to be used to describe exercise treatments.
| Exercise characteristic | Characteristic description or guidance |
| Exercise specificity | The focus of the exercises: whole body (many muscle groups from around the body, generally seeking to improve overall fitness), back‐specific (exercises concentrated around muscle groups that support the back), or both |
| Program design | The level of individualisation of exercise program: individually designed (specifically designed for the individual participant), standard design (fixed exercise program, which may vary slightly in intensity, frequency, and/or duration) |
| Program delivery mode | Delivery mode of the exercise treatment: independent (home exercise), independent with therapist follow‐up, group, group with therapist follow‐up, or individual (1:1) with therapist |
| Dosage | How much exercise did participants receive, considering length of sessions, number of sessions, programme duration, and adherence or completion |
| Delivering therapist/ provider | Therapist delivering the exercise treatment: healthcare professional (i.e. physical therapist, physiotherapist, chiropractor, other healthcare professional), exercise specialist (e.g. yoga, Pilates instructor), other |
| Graded activity approach | If the exercise progressed using a time‐contingent manner, regardless of pain, with use of quotas and pacing |
| Co‐interventions provided | All co‐interventions delivered to the exercise treatment group will be recorded, including, but not limited to, education, psychological therapy, mind‐body approach, medications, or manual therapies |
In the overarching review, we will include studies that compared exercise treatment to (1) placebo/sham/attention control, (2) no trial treatment, (3) other conservative therapy, or (4) another exercise group (Table 3). We will classify comparisons as no trial treatment, when the comparison group regimen is described as no specific treatment provided by the trial, including waiting list (no other description provided by the trial authors), control group (described specifically as no intervention; or no other description provided by the trial authors), usual/normal care (stated that participants could receive normal care, but this was not controlled by trial, and may be offered to all groups), or exercise and comparison groups are offered, or receive, the same co‐interventions, allowing the effect of the exercise treatment to be isolated. We will categorise a comparison as other conservative therapy, when participants received any non‐exercise interventions (categorised as education, education with self‐management support, psychological therapy, manual therapy, electrotherapy, relaxation, physical therapy (non‐exercise), acupuncture, medication, back school); and another exercise group, when the comparison included another eligible exercise intervention that differed in type or design/delivery approach. We will exclude studies that compare exercise treatment with spinal surgery.
3. Comparison type, categories, and descriptions.
| Category for analysis | Category extracted | Description or guidance |
| Placebo/sham | Placebo/sham | The group is described as a sham, placebo, or attention control, or is judged to be intended by the trial authors to be a sham, placebo, or attention control. Participants unknowingly participated in a placebo or sham treatment that was feasibly effective. |
| No trial treatment | Wait list | Participants received no treatment but a wait list. |
| Control group | The group is described specifically as no intervention or control group; or no other description is provided by the trial authors. | |
| Usual/normal care | It is stated that participants could receive normal care; but this was not controlled by the trial and may be offered to all treatment groups. | |
| Same co‐interventions | The exercise and comparison groups are offered, or receive, the same co‐interventions, allowing the effect of the exercise treatment to be isolated. | |
| Other conservative treatment | Participants received an ‘other conservative treatment’ (see Table 4). | |
| Not specified | No details about treatment are provided. | |
| Other | Comparison type not described by options above; groups that are specifically instructed or monitored (or both) as receiving no treatment will be recorded here and described separately. | |
5. Other conservative treatment categories and descriptions.
| Category for analysis | Category extracted | Description or guidance |
| Education | Advice to stay active | Healthcare professional or study staff telling participants to keep moving (no bed rest) to help deal with back pain symptoms. |
| Back Book | Written educational material provided with a goal to improve an individual’s understanding of their back problems and what they should do about them; provided in the form of a booklet | |
| Ergonomic training | Education about posture and environmental design factors (often workplace) | |
| Education alone | Information on conditions to enhance the participants’ factual knowledge and conceptual understanding of mechanisms related to health maintenance and improvements | |
| Education with self‐management support | Education with self‐management support | Treatments to improve the individual's ability to manage the symptoms, treatment, physical and psychosocial consequences, or lifestyle changes (or a combination) inherent in living with a chronic condition; often includes behaviour change techniques |
| Psychological therapy | Psychological therapy or counselling | Cognitive behavioural therapy, psychiatry, stress management, etc. |
| Manual therapy | Massage | Soft‐tissue manipulation, using the hands or a mechanical device |
| Mobilisation/manipulation (passive) | Mobilisation or manipulation techniques administered by a therapist | |
| Traction | Traction of the lumbar region including mechanical or motorised traction, manual, or auto‐traction | |
| Trigger point therapy | Treatment that involves the manual application of pressure to trigger points | |
| Electrotherapy | Electrotherapy | Participants received ultrasonography, transcutaneous electrical nerve stimulation (TENS), percutaneous electrical nerve stimulation (PENS), laser therapy |
| Relaxation | Balneotherapy | Treatment that involves bathing in water, including spa therapy or mineral baths |
| Meditation | Meditation treatment can be done in various postures and ways, including focusing on breathing, on an object, or repeating words (a mantra). | |
| Breathing practice | Relaxation technique focused on breathing | |
| Physical therapy (non‐exercise) | Heat | Superficial heat treatments including, for example, hot water bottles, hot towels, heat pads, or wraps |
| Ice | Superficial cold treatments including, for example, ice, cold towels, or cold gel packs | |
| Physical therapy (non‐exercise mixed) | Multiple non‐exercise physical therapy modalities (e.g. heat and TENS) | |
| Brace/lumbar support | Taping, brace, or an implement used on furniture to provide postural correction | |
| Alexander Technique | A taught treatment involving a series of movements designed to correct posture and bring the body into natural alignment | |
| Acupuncture | Acupuncture (including electroacupuncture) | Treatment involving needling, including traditional acupuncture, or dry‐needling |
| Medication | Medication (e.g. NSAIDS, analgesics, steroids) | Steroids, analgesics, non‐steroidal anti‐inflammatory drugs, etc. |
| Back School | Back School | A treatment program given to groups of people that includes education on biomechanics, posture, ergonomics, and back exercises |
NSAID: non‐steroidal anti‐inflammatory drug
Types of outcome measures
Major outcomes
We will fully report the following major outcomes when they are available (prioritised measures described in brackets).
Pain intensity (visual analogue scale (VAS); numeric rating scale (NRS (Melzack 1975)).
Functional limitations (Roland‐Morris Disability Questionnaire (RMDQ (Roland 2000)); Oswestry Disability Index (ODI (Fairbank 1980))).
Health‐related quality of life (SF‐36 or SF‐12 (36/12‐Item Short Form Health Survey (Ware 1992)), PROMIS‐GH‐10 (10‐item Patient‐Reported Outcomes Measurement Information System Global Health short form (Hays 2009)), EQ‐5D (EuroQoL 5 Domain (EuroQol 2019)), CDC HRQOL‐14 (or other versions (NCCDPHP 1993)), WHOQOL‐BREF (World Health Organization Quality of Life Scale (WHOQOL Group 1998)), NHP (Nottingham Health Profile (Wiklund 1990)), QOLS (Quality of Life Scale (Burckhardt 2003)), or SIP (Sickness Impact Profile (Bergner 1976)).
Psychological functioning (Beck Depression Inventory (Beck 1987; Beck 1988)); Zung Depression Index (Zung 1986); Patient Health Questionnaire‐9 (PHQ‐9 (Kroenke 2001)); Montgomery‐Asberg Depression Rating Scale (MADRS (Davidson 1986)); Hamilton Rating Scale for Depression (HRSD (Hamilton 1986)); Center for Epidemiologic Studies Depression Scale (CES‐D (Radloff 1977)); Hospital Anxiety and Depression Scale (HADS (Zigmond 1983)); Hopkins Symptoms Checklist for anxiety and depression (HSCL (Derogatis 1974))
Adverse events
We selected major outcome domains and measurements that include recommendations from the core outcome set for low back pain clinical trials (Chiarotto 2018).
Minor outcomes
We will also describe measurement for the following minor outcomes, as reported in the trial publications (minor outcomes will not be synthesised in this Cochrane Review).
Return to work/absenteeism (subjective or objective assessment of work absence or sick leave, or return to work rates for individuals absent from work at baseline)
Global improvement or perceived recovery (any measure of participant‐reported recovery, resolution, or improvement, as defined by the trial)
Satisfaction (any measure of participant‐reported satisfaction with low back pain, or treatment outcomes (or both), as defined by the trial)
Medication use (any measure of presence/absence, or count of pain‐relieving medication use, as defined by the trial)
Self‐efficacy (any measure of participant‐reported confidence to perform activities and achieve goals despite low back pain symptoms)
Cost (any measurement of healthcare expenditure reported to be available in the trial)
Timing of outcome assessments
We will record all outcomes and time periods available and reported, and extract outcome data for major outcomes. We will define specific follow‐up periods as: end of treatment (follow‐up measured within one week before or after the end of treatment), moderate‐term post‐treatment (follow‐up measured at more than 14 weeks to 47 weeks after the end of treatment, closest to 6 months) and long‐term post‐treatment (follow‐up measured at least 48 weeks after the end of treatment, closest to 12 months).
For our primary analyses we will use study data from the follow‐up period closest to end of treatment. To accommodate variations in reporting, and allow the largest number of sufficiently similar eligible studies, we will define closest to end of treatment as the available follow‐up period that is at least half‐way through the treatment program, and up to 14 weeks post‐treatment. In applying this definition, we will prioritise time periods available for a trial in this order: (1) follow‐up measured zero to six weeks post‐treatment, (2) follow‐up measured more than half‐way through the treatment to the end of treatment, then (3) follow‐up measured between 6 weeks and 14 weeks post‐treatment.
Minimally important differences
We will interpret a clinically important difference in results for exercise treatment compared to comparison treatments considering cost, inconvenience, potential harms, and (assumed) participant preference. For all outcomes, we will initially consider a difference of 10% of the scale range for comparisons of exercise treatment to no treatment. This reflects a difference in pain and functional limitations of 10 points out of 100, and is compatible with the smallest worthwhile effect, based on a 20% reduction, the estimated median for the smallest worthwhile self‐reported effect with physiotherapy treatment compared to without physiotherapy (Ferreira 2013). We found this in the previous update of this review, in which the average baseline pain in the included studies was 50.9 (95% confidence interval (CI) 49.1 to 52.8), and the average baseline functional limitation was 38.9 (95% CI 35.8 to 42.0).
We will interpret smaller differences in effectiveness of exercise treatment compared to other conservative treatments as probably meaningful, if the 95% CI is entirely on one side of the line of no effect. We consider this to be appropriate, given similar inconveniences and adverse effects for comparison treatments considered in this review (Qaseem 2017). We will consider differences to be statistically significant at the 5% level.
For our network meta‐analyses, we will define five categories of results (categories 1 to 4 favouring one treatment) as we did previously (Hayden 2021b): (1) clinically important difference (mean effect size ≥ 10 absolute; both 95% CI in the same direction); (2) moderate difference compatible with a clinically important difference (mean effect size < 10 absolute; highest 95% CI ≥ 10, both 95% CI in the same direction); (3) moderate difference (mean effect size < 10 absolute; highest 95% CI ≥ 10, opposite direction 95% CI < 2.5, or highest 95% CI ≥ 5, both 95% CI in the same direction); (4) small difference (mean effect size < 10 absolute; highest 95% CI ≥ 5, opposite direction 95% CI < 2.5, or highest 95% CI < 5, both 95% CI in the same direction, or highest 95% CI ≥ 10, opposite direction 95% CI >2.5); and (5) no difference.
Search methods for identification of studies
Electronic searches
We ran updated and optimised electronic searches for this systematic review to ensure adequate retrieval for the broad overarching review, as well as the embedded focused Cochrane Reviews (sub‐reviews). See Appendix 1 for the full electronic strategy. The search optimisation was completed by LB in November 2021, and resulted in us including additional exercise type terms, and searching fewer databases (Search optimisation report: Appendix 2). A second, independent health librarian reviewed the search strategy, using the PRESS Checklist (McGowan 2016).
We ran the full electronic search strategy on 17 May 2022. We reconciled the results against the search return of the previous Cochrane Review (Hayden 2021a).
We searched the following databases with no date or language restrictions:
Cochrane Central Register of Controlled Trials (CENTRAL; 2022, Issue 5) in the Cochrane Library (searched 17 May 2022);
MEDLINE OvidSP (1946 to 17 May 2022);
Embase (Embase.com; 1974 to 17 May 2022).
We are managing the citations with EndNote X8 software (Endnote 2017). We will use the same search strategy for related sub‐reviews.
Searching other resources
Many systematic reviews on the effectiveness of exercise treatment for low back pain have been published. We will identify relevant systematic reviews in the electronic search and from team records; we will review the included studies lists from these systematic reviews for other potentially relevant trials.
We will review trial protocols identified and registrations identified by the CENTRAL search, which includes ClinicalTrials.gov (www.ClinicalTrials.gov) and the World Health Organization International Clinical Trials Registry Platform (ICTRP) Search Portal (apps.who.int/trialsearch/). We will contact the authors of all ongoing trials via an emailed REDCap survey to determine if trials are (a) complete, and (b) published (Harris 2019).
Prior to analyses and reporting, we will search for retractions and publication corrections within our set of eligible studies, using the software Zotero, which is integrated with the Retraction Watch database (Retraction Watch 2023).
Data collection and analysis
We will follow structured procedures for citation management, study selection, and data extraction, facilitated by evidence synthesis tools and online software. We will prescreen new search results using Cochrane Screen4Me resources (Noel‐Storr 2021), including a machine‐learning algorithm (Thomas 2021), and Cochrane Crowd RCT classification (Noel‐Storr 2021). After study screening, we will complete data extraction and risk of bias assessment, facilitated by a web‐based electronic systematic review software, Distiller SR (Distiller SR). We will record all data on forms modified from those used in our previous review, and pilot tested with collaborative review team members. All collaborative review team members contributing to study selection and data extraction will complete training/consensus exercises before the project start to ensure reliability between reviewers.
Selection of studies
Pairs of independent reviewers from our collaborative review pool of 35 reviewers will screen citations based on the title and abstract, and subsequently full text, for inclusion in the review. Consensus, and if necessary, a third reviewer will be used to resolve disagreements. We will assess selection criteria to judge likely eligibility of each citation; citations that remain unclear after review of the title and abstract will proceed to full‐text assessment. At the full‐text assessment level, we will contact the trial authors when an inclusion criterion remains unclear. We will exclude conference proceedings, theses, opinion pieces, correspondence, and stand‐alone abstracts. We will assess studies published in languages other than English for inclusion and include them in the review whenever feasible, using English language abstracts, translation tools, and review by review team members and colleagues familiar with the language of publication.
We will record the selection process in sufficient detail to complete a PRISMA flow diagram and characteristics of excluded studies table (Page 2021).
We will use a systematic process to identify and exclude duplicates, and collate multiple reports of the same study. We will link studies by searching across identified studies for overlapping author names, trial registration number, and baseline sample size. If matches are identified, we will check the full text to confirm the sample and intervention descriptions. We will link multiple reports of the same study under a single reference ID so that each study, rather than each report, is the unit of interest in the review.
Data extraction and management
For each trial, a single reviewer will extract study information and a second reviewer will check all extracted data points against the original study publication. Reviewers will not be blinded to authors, institution, or journal of publication, as this is not feasible with our resources, and because the collaborative review team is very familiar with the literature.
We have developed clear guidance documentation for all study characteristics extracted, which corresponds with detailed extraction forms (Appendix 3). We will extract characteristics of the study methods, population (participant population source and setting, study inclusion criteria, mean duration of pain, symptom characteristics, mean age, and sex of participant populations), exercise intervention (description and types of exercise treatment, duration and number of treatment sessions, intervention delivery type, and any additional interventions according to characteristics reported in the CERT guidelines (Slade 2016), as required by Cochrane Musculoskeletal), comparison group (category and main type for the other conservative treatments), and outcome data. Study characteristics extracted as potential treatment effect modifiers (to assess the transitivity assumption for our network meta‐analyses) will be: participant recruitment setting, mean age, duration of back pain, leg pain or neurologic symptoms, and baseline pain and functional limitations. Detailed description of the exercise characteristics extracted is provided in Table 1. We will extract results for major continuous outcomes as final value scores, when available, for inclusion in meta‐analyses. For dichotomous adverse event outcomes, we will extract information about how adverse events were measured, the number of events, and the number of participants per treatment group, if available. For any cross‐over trials, we will extract outcome data only from the follow‐up periods before the cross‐over, to avoid period and carryover effects; if data before the cross‐over period are not available, we will consider the outcome data missing.
We will contact trial authors to request any missing study characteristics, data points, or risk of bias information, when required. We will email them, and include a link to a REDCap data capture form, where both extracted data and missing fields are clearly displayed (Harris 2019). We will ask the authors to complete the missing fields, correct any incorrect extracted data, and then submit the survey through REDCap. This survey approach for data checking was used successfully in the previous update of this review, and resulted in a much higher response rate compared to emailing clarifying questions with no survey (24% versus 47% (Hayden 2021a)).
We will use robust methods for data collection, including comprehensive training, clear data collection forms, and guidance documentation. However, we still recognise that incorrect values in systematic review data are common (including errors in data extraction, misunderstanding by reviewers, or misreporting by trialists). We will follow predefined guidelines for data validation and cleaning to identify and correct faulty data, including before and after electronic transfer of data to the RevMan Web file (RevMan Web). This will include inspection, verification, cleaning, re‐verification, and reporting (Appendix 4).
Assessment of risk of bias in included studies
‐We will conduct the risk of bias assessment for randomised trials using a modified version of the RoB 1 criteria recommended by Cochrane (Higgins 2011), and Cochrane Back and Neck (Furlan 2015), which includes twelve items (Table 5 plus one ‘other’ biases item). We will classify individual items as high risk, low risk, or unclear risk. Two review authors will independently conduct risk of bias assessments, with consensus. Any disagreements that cannot be resolved through discussion will be referred to a third reviewer. We will calculate inter‐rater reliability, related to risk of bias, as overall agreement and Kappa scores, based on assessments before consensus judgements are reached.
4. Items assessed as low, high, unclear risk of bias, modified from Cochrane Back & Neck Methods Guidelines.
| ROB 1 item | Description | Judgement for low ROB* |
| 1. Method of randomisation | Describe the method used to generate the allocation sequence in sufficient detail to allow an assessment of whether it should produce comparable groups. | A random (unpredictable) assignment sequence. Examples of adequate methods are coin toss (for studies with 2 groups), rolling a dice (for studies with 2 or more groups), drawing of balls of different colours, drawing of ballots with the study group labels from a dark bag, computer‐generated random sequence, pre‐ordered sealed envelopes, sequentially‐ordered vials, telephone call to a central office, and pre‐ordered list of treatment assignments. Examples of inadequate methods are: alternation, birthdate, social insurance/security number, date in which they are invited to participate in the study, and hospital registration number. |
| 2. Treatment allocation | Describe the method used to conceal the allocation sequence in sufficient detail to determine whether intervention allocations could have been foreseen in advance of, or during, enrolment. | Assignment generated by an independent person not responsible for determining the eligibility of the participants. This person has no information about the persons included in the trial, and has no influence on the assignment sequence or on the decision about eligibility of the person. |
| 3. Participant blinding | Describe all measures used, if any, to blind study participants from knowledge of which intervention a participant received. Provide any information relating to whether the intended blinding was effective. | Index and comparison groups are indistinguishable for the participants; or the success of blinding was tested among the participants, and it was successful (i.e. participants in both the index and comparison groups felt that they received ‘the best’ treatment); or any lack of blinding did not lead to deviations from the intended intervention. |
| 4. Care provider blinding | Describe all measures used, if any, to blind personnel from knowledge of which intervention a participant received. Provide any information relating to whether the intended blinding was effective. | Index and comparison groups are indistinguishable for the care providers; or the success of blinding was tested among the care providers, and it was successful (i.e. care providers for both the index and comparison groups felt that they were delivering ‘the best’ treatment); or any lack of blinding did not lead to deviations from the intended intervention. |
| 5. Outcome assessor blinding | Describe all measures used, if any, to blind outcome assessors from knowledge of which intervention a participant received. Provide any information relating to whether the intended blinding was effective. | Adequacy of blinding should be assessed for each primary outcome separately. This item should be scored low risk if the success of blinding was tested among the outcome assessors, and it was successful, or:
|
| 6. Dropout rate | Describe the completeness of outcome data for each main outcome, including attrition and exclusions from the analysis. State whether attrition and exclusions were reported, the numbers in each intervention group (compared with total randomised participants), reasons for attrition/exclusions where reported, and any re‐inclusions in analyses performed by the review authors. | The number of participants who were included in the study but did not complete the observation period, or who were not included in the analysis, must be described and reasons given. If the percentage of withdrawals and dropouts does not exceed 20% for short‐term follow‐up and 30% for long‐term follow‐up, and does not lead to substantial bias, a low risk is scored. (N.B. these percentages are arbitrary, not supported by literature). |
| 7. Intention‐to‐treat | Describe the intention‐to‐treat analysis. | All randomised participants are reported/analysed in the group to which they were allocated by randomisation for the most important outcome follow‐up time points (minus missing values) irrespective of noncompliance and co‐interventions. |
| 8. Selective reporting | State how the possibility of selective outcome reporting was examined by the review authors, and what was found. | All results from all prespecified outcomes have been adequately reported in the published report of the trial. This information is either obtained by comparing the protocol and the report, or in the absence of the protocol, assessing that the published report includes enough information to make this judgement. |
| 9. Baseline characteristics | Describe the similarity of groups on key characteristics at baseline. | Groups have to be similar at baseline regarding demographic factors, duration and severity of complaints, percentage of participants with neurological symptoms, and value of main outcome measure(s). |
| 10. Co‐interventions | Describe if co‐interventions were avoided or were comparable. | If there were no co‐interventions, or they were similar between the index and control groups |
| 11. Compliance | Describe compliance rates across groups. | The reviewer determines if the compliance with the interventions is acceptable, based on the reported intensity, duration, number and frequency of sessions for both the index intervention and control intervention(s). For example, physiotherapy treatment is usually administered for several sessions; therefore, it is necessary to assess how many sessions each participant attended. For single session interventions (e.g. surgery), this item is irrelevant (i.e. low risk). |
| 12. Timing of outcomes | Describe the similarity in timing of the outcome assessment of the groups. | Timing of outcome assessment should be identical for all intervention groups for the major outcomes being assessed. |
*modified from Cochrane Back & Neck Methods Guideline (Furlan 2015)
We will use the twelve modified RoB 1 items to assess the following bias domains.
Selection bias (method of randomisation, treatment allocation concealment, similarity of baseline characteristics)
Performance bias (blinding of participants and care provider, intention‐to‐treat)
Attrition bias (missing outcome data/dropouts)
Detection bias (blinding of outcome assessors, similar timing of outcome assessment)
Reporting bias (selective outcome reporting)
Other biases (avoidance of co‐interventions, compliance).
For any cluster‐randomised trials, we will assess an additional domain, bias arising from the timing and recruitment of participants. We will report judgements for each bias domain in a risk of bias table. We will assess studies as having an overall high risk of bias if any domain is judged to have high risk of bias. We will assess risk of bias for all major self‐reported outcomes combined; if judgements differ according to major outcome, we will assess the items separately, and base our domain and overall assessment on the outcome judged to have the highest risk of bias. We will use sensitivity analyses to assess the robustness of our findings, by excluding from the syntheses any studies considered to be at high risk of bias.
In addition to risk of bias, we will also comprehensively assess several characteristics of the included studies related to integrity of the research. In our 2021 review, we assessed integrity of the research using the following criteria: evidence of plagiarism, publication in a presumed predatory journal, high risk of bias, and inadequate reporting of minimal basic CONSORT items (Hayden 2020; Hayden 2021c). In this review, we will use a similar multicomponent assessment, with some modifications to criteria used: we will include assessment of prospective trial registration, we will assess adequate reporting more comprehensively, we will not use risk of bias assessment, and we will not conduct a plagiarism assessment, as this criterion is time‐intensive and requires specialised software. We will classify a trial as prospectively registered if the trial was registered before or within a month of the trial start date (date that the first participant was enroled). We will classify a trial as retrospectively registered if the trial was registered more than one month after the trial start date (e.g. trial registered in August 2010 or later for a trial that began July 1, 2010). Two reviewers will independently assess trials as being published in a presumed predatory journal using the decision algorithm of Boulos and colleagues (Boulos 2022). We will comprehensively assess reporting for eligible trials by assessing 24 items related to study design description (7 items: randomisation approach; recruitment dates; author contact information; research ethics board review; conflict of interest statement; funding source for trial; complete flow chart or details about participant flow), population description (8 items: country of study conduct; inclusion criteria for pain duration; pain duration of population; inclusion criteria for back pain type; population setting or source; participant sex; participant age; baseline pain or function), exercise treatment description (7 items: programme individualisation; programme delivery site; programme delivery mode; programme dosage; number of sessions; programme duration; programme adherence), outcome reporting (2 items: pain, function or health‐related quality of life outcome; adverse effects outcome). We will judge that a study is inadequately reported if the study reports less than 75% of required items (17 or fewer of the 24 items). We will exclude studies from our review if they: (1) are not prospectively registered, and (2) are published in a presumed predatory journal, and (3) are judged to have inadequate reporting.
Assessment of bias in conducting the systematic review
We will conduct the review according to this protocol and report any deviations from it in the Differences between protocol and review section of the systematic review.
Measures of treatment effect
Our primary meta‐analyses will assess treatment effects on pain, functional limitations, health‐related quality of life, and psychological functioning (core outcomes reliably reported in eligible trials).
We will synthesise pain intensity measured by a pain scale, prioritising VAS, then NRS. Research has established agreement of the VAS and NRS pain scales, and this same research established that they can be used interchangeably to assess back pain intensity (Shafshak 2021).
We will synthesise functional limitations measured with a back pain‐specific scale, for example, the RMDQ or the ODI, preferentially selecting the RMDQ when it is available. There is evidence that the RMDQ and the ODI are highly correlated, and similarly responsive enough to combine in meta‐analysis (Chiarotto 2016).
We will re‐scale the individual trial outcomes for pain and functioning to 0 to 100 points for meta‐analyses (for example, a VAS pain score (standard deviation) of 5.1 (2.3) out of 10 would be re‐scaled to 51 (23) out of 100), where a negative mean difference indicates improvement (i.e. decreased pain and decreased functional limitations). Re‐scaling is accepted and common in the back pain field, and facilitates comparison and interpretability of the syntheses (Kopec 2000).
We will synthesise health‐related quality of life outcomes using one of a prioritised list of measures, as available: SF‐36 or SF‐12, PROMIS‐GH‐10, EQ‐5D, CDC HRQOL‐14 (or other versions), WHOQOL‐BREF, NHP, QOLS, or SIP. For the SF‐36, SF‐12, or PROMIS‐GH‐10, we will separately synthesise the physical and mental component scores, as recommended.
We will synthesise psychological functioning as depression outcomes, using one of a prioritised list of measures, as available: Beck Depression Inventory, Zung Depression Index, PHQ‐9, MADRS, HRSD, CES‐D, HADS, or HSCL.
We will analyse our continuous major outcomes as mean difference (MD) and 95% CIs, when a sufficiently similar scale is used to measure an outcome. We will enter data presented as a scale with a consistent direction of effect across studies.
When sufficiently different scales are used to measure the same conceptual outcome (e.g. depression), we will calculate the standardised mean difference (SMD) instead, with corresponding 95% CIs. SMDs will be back‐translated to a scale of 0 to 100 by multiplying the SMD by a typical among‐person standard deviation, estimated as the standard deviation of the control groups at baseline from the most representative trials (Higgins 2022b).
If synthesis is possible, we will analyse our dichotomous major outcome, adverse events, as odds ratios with 95% CIs.
Unit of analysis issues
Opportunities for unit of analysis issues in this systematic review are due to: (1) repeated observations on participants, (2) studies with more than two intervention groups, and (3) non‐parallel study designs (cross‐over and cluster‐randomised trials).
We will assess available data from multiple follow‐up periods of the same treatment groups, by defining different outcomes based on different periods of follow‐up: end of treatment, moderate‐term post‐treatment, and long‐term post‐treatment. We will use data from the follow‐up period closest to end of treatment for all primary analyses.
Exercise treatment groups from included trials will be included in the pairwise meta‐analyses if they have an independent no treatment or other conservative treatment comparison group (multiple pairwise comparisons may be available in a trial). This requirement will mean that we will not include studies with no comparison group (i.e. trials that contrasted multiple exercise treatment groups only) in the pairwise meta‐analyses, and we will not double count comparison groups in the meta‐analyses. This latter criterion is necessary to avoid correlation in effect sizes resulting from the use of repeated comparison data. We will select groups to include in the analyses based on relevance to the review selection criteria. If a single relevant comparison group exists in a trial with multiple exercise groups, we will split the shared comparison group into two or more groups with smaller sample sizes to allow inclusion in meta‐analyses; this will allow investigation of heterogeneity across treatment arms (Higgins 2022c).
Our additional network meta‐analysis and meta‐regression analyses will use available data from all treatment groups, including indirect and direct comparisons, and allow the inclusion of data from studies with only multiple exercise group comparisons.
For any eligible cross‐over trials, we will only extract and analyse outcome data from the follow‐up periods before the cross‐over to avoid period and carryover effects. For cluster‐randomised trials, we will assess the appropriateness of the analyses to control for correlation of individuals within clusters. If clustering is not accounted for in a trial, we will reduce the sample size to an estimate of the effective sample size for continuous outcome data to down‐weight the study and limit potential unit of analysis error (Higgins 2022a).
Dealing with missing data
We will request any missing data from the trial reports from the individual study authors, first. If we are unable to contact the original authors, or they are unable to supply missing data, we will impute missing variance scores, using the mean variance from studies that are not at high risk of bias and have similar populations of low back pain. When data in study papers are reported as a median and interquartile range (IQR), we will use the median to estimate the mean for studies with moderate to large sample sizes (N > 25); for studies with small sample sizes, we will use the formula proposed by Hozo and colleagues (Hozo 2005). We will calculate the standard deviation with the width of the IQR equivalent to 1.35 times the standard deviation (Higgins 2002). We will treat the standard deviation as missing in studies where only a range is presented along with the median. We will conduct sensitivity analyses to explore the impact of assumptions about missing or incomplete data from the trial reports.
Assessment of heterogeneity
We will pool data if we judge studies to be clinically homogeneous with regard to study population, intervention, and outcomes. We will consider clinical heterogeneity and inspect forest plots to assess the direction and size of effects, and the degree of overlap between confidence intervals.
We will use the I² statistic to quantify inconsistency among the trials in each analysis (Chi², P < 0.1). If we identify substantial heterogeneity, we will report it and explore possible causes by prespecified subgroup analysis. We will use this approximate guide for the interpretation of an I² value (Deeks 2022).
0% to 40% might not be important
30% to 60% may represent moderate heterogeneity
50% to 90% may represent substantial heterogeneity
75% to 100% represents considerable heterogeneity
These overlapping intervals reflect that the interpretation of the I2 statistic depends on the value as well as the size and direction of the treatment effect and variance of the I2 estimate. The assessment of heterogeneity will inform our appraisal of the certainty of evidence available using the GRADE framework.
Assessment of reporting biases
We will evaluate between‐study publication bias for pairwise comparisons, with Egger's test and funnel plots (Egger 1997). We will also assess within‐study publication bias by considering whether all expected outcomes were reported for each follow‐up period of interest. If we are suspicious of reporting bias for a particular study, we will contact study authors for additional information, and attempt to locate the study protocol or trial registration to determine if there were differences between the protocol and publication.
Data synthesis
The primary analyses will include descriptive analyses, pairwise meta‐analyses, and network meta‐analyses of all eligible studies. We will use two network meta‐analysis approaches, standard network meta‐analysis to estimate the treatment effects for specific exercise types, and component network meta‐analysis to investigate the contribution of exercise treatment components.
We will conduct our pairwise and standard network meta‐analyses, subgroup, and sensitivity analyses for all major outcomes, using the time period closest to the end of treatment. We will conduct component network meta‐analyses for pain and functional limitations outcomes only.
Pairwise meta‐analyses
We will construct forest plots for all pairwise comparisons in Review Manager Web, to give a pictorial overview of all study results for each major outcome (RevMan Web). We will use pairwise meta‐analyses, conducted for continuous outcome measures, by pooling mean differences with random‐effects models and data from at least two studies (DerSimonian 1986). We will calculate prediction intervals to estimate the likely effect of exercise treatment in individual study settings (Riley 2011).
Standard network meta‐analyses
We will conduct standard network meta‐analyses to estimate the effects and associated uncertainty of the interventions on each major outcome separately. Treatment nodes will include aerobic, core strengthening, general strength training, McKenzie therapy, motor control exercise, Pilates, Qigong, stretching or flexibility/mobilising exercises, Tai Chi, yoga, mixed exercise, plus three comparison categories: placebo/sham, no trial treatment, and other conservative treatment.
For standard network meta‐analyses, we will fit a frequentist inconsistency model, using contrast‐based linear mixed‐effects modelling. The models will include fixed‐effect parameters for the effects of intervention, baseline outcome value (i.e. pain outcome models adjusting for pain at baseline; functional limitations outcome models adjusting for baseline functional limitations), exercise dose and additional conservative treatments, and their interactions; and random‐effects terms to account for correlation between observed effects in trials with more than two groups. We will specify random‐effects terms, using a compound symmetric covariance structure with rho = 0.5 (Higgins 1996).
Our primary network meta‐analyses will not produce ranking of exercise treatments, due to limitations and potential misinterpretation of ranking results. Alternatively, we will report estimates of the relative treatment effects, with uncertainty of estimates, and interpret results by considering compatibility with clinically important differences. For completeness, we will present ranks for each outcome as a secondary output. In these analyses, we will identify the exercise treatment most likely to have the largest mean outcome improvement (mean difference using ranking metric probability of best value), and we will identify the treatment most likely to produce clinically important improvement (absolute probability using ranking metric median rank).
Component network meta‐analyses
We will extend the standard network meta‐analysis approach to investigate the effects and associated uncertainty for specific treatment components, individually and in combination, by deconstructing the exercise treatment programs and conservative treatments, including: four classes of exercise treatment types (cardiorespiratory/aerobic, resistance, flexibility, and neuromotor training); program design (individually designed versus standard); delivery type (independent, independent with follow‐up, group‐supervised with/without follow up, individually‐supervised); dose (high versus low dose); and other conservative treatments (education, education with self‐management support, psychological therapy, manual therapy, electrotherapy, relaxation, physical therapy (non‐exercise), acupuncture, medication, back school). We will define a no treatment comparison for these analyses including placebo/sham or no trial treatment comparisons. We will conduct this exploratory analysis for pain and functional limitations outcomes only, as we expect an insufficient number of trials will report other major outcomes, resulting in a sparse network.
For component network meta‐analyses, we plan to fit two models: (1) additive main‐effects model (investigating the effect for each treatment component individually and in combination), and (2) full interaction model (if possible, investigating each combination of treatment components).
We will use R packages for descriptive analyses, data checking, production of funnel plots, pairwise, network meta‐analyses, component network meta‐analyses and related graphical presentations, and calculation of prediction intervals (R Foundation for Statistical Computing), and RevMan Web for pairwise meta‐analyses and data presentation (RevMan Web).
Subgroup analysis and investigation of heterogeneity
We will use subgroup analyses for all major outcomes, to explore heterogeneity due to study‐level variables, such as population source (healthcare, occupational, general or mixed population) and duration of chronic low back pain (moderate‐, long‐term). We will also explore specific types of conservative treatment comparisons, including education, education with self‐management support, psychological therapy, manual therapy, electrotherapy, relaxation, physical therapy (non‐exercise), acupuncture, medication, and back school. We will compare major outcomes for subgroups of studies (Song 2003).
Assessment of network transitivity
We will assume that all participants in the included trials were equally likely to be randomised to any of the interventions in the included trials (i.e. that the transitivity assumption was plausible). Nonetheless, we will describe population variables that may act as treatment effect modifiers, and assess their distribution across network comparisons: participant setting, mean age, duration of low back pain, leg pain or neurologic symptoms, and baseline pain and functional limitation outcome values. If there are distributional differences across comparisons in the investigated treatment effect modifiers for a network meta‐analysis model, then we will conduct a secondary analysis adjusting for these additional characteristics.
Assessment of incoherence
We will evaluate incoherence (i.e. agreement between direct and indirect evidence) of the major outcome networks globally and evaluate local incoherence for each treatment comparison using side‐splitting (Dias 2010), and by evaluating statistical incoherence of the network separately in every closed loop (Chaimani 2015). Local incoherence will be considered statistically significant if loop‐specific 95% confidence intervals do not include zero.
If the transitivity assumption seems implausible for any model, based on the clinical and statistical assessments above, then we will abandon the planned network meta‐analysis model and report the synthesis using direct evidence only.
Sensitivity analysis
We will conduct sensitivity analyses for all major outcomes to explore the impact of methodological decisions, and to further explore residual heterogeneity (and incoherence for network meta‐analyses) in the primary analyses. We will conduct sensitivity analyses excluding studies with high risk of bias, with no prospective trial registration, outliers in our closest to the end of treatment follow‐up period, and to test assumptions about imputing data for studies that did not adequately present variance scores and where median values were reported. We will conduct sensitivity analyses that omit study outcome data from meta‐analyses that were judged to be outliers to explore the impact of extreme study results on review conclusions. A study will be judged to have an outlying (or improbable) mean outcome for pain and functional limitation outcomes if the absolute difference between any exercise group and any comparison group over all available follow‐ups was greater than a predetermined threshold of 30/100 for pain and 20/100 for functional limitations, selected based on clinical judgement (we will not consider other major outcomes in this judgement).
We will assess changes in interpretation of results for sensitivity analyses with an algorithm to identify changes in the interpretation of the effect direction, size, and compatibility with a clinically important difference. We will define important change in the interpretation of results, from primary analysis to sensitivity analyses, based on the number of changes in the clinical importance results category (0 to 1 category changes = no concerns; 2 to 3 category changes = some concerns; and 4+ category changes = major concerns).
Summary of findings and assessment of the certainty of the evidence
We will present results for our major outcomes of pain, functional limitations, and adverse events for pairwise comparisons in the summary of findings tables, guided by the GRADE framework. For pairwise meta‐analyses, we will present the tables for placebo/sham and no treatment comparisons. These tables include a summary of the number of studies included in the review, continuous outcome measure results reflecting absolute treatment effects, and a statement regarding the overall quality of the evidence available.
We will use the CINeMA web application (which adapts GRADE domains for network meta‐analysis) to evaluate confidence in findings from the network meta‐analysis due to: risk of bias within comparisons, publication bias, indirectness, imprecision, heterogeneity, and incoherence (Nikolakopoulou 2020). We will present our main review results from standard network meta‐analyses using a summary of findings table layout, modified from recommendations of Yepes‐Nunez and colleagues (Yepes‐Nuñez 2019). We will interpret and report the findings from our network meta‐analyses using a partially contextualised approach (Brignardello‐Petersen 2017), and cautiously, considering results to be more certain if they behave as expected by theoretical mechanisms, and across multiple comparisons and models (e.g. results for different outcomes, planned subgroup and sensitivity analyses).
Acknowledgements
The Canadian Institutes of Health Research provides funding for this project (Project Grant Competition, PJT‐173478). We appreciate insights and contributions of Kristy Hancock, Evidence Synthesis Co‐ordinator at the Maritime SPOR Support Unit for assistance with the literature search, and additional central and subreview team members contributing to study conduct: Nora Bakaa, Jennifer Cartwright, Pedro Isaac Santos Chaves, Gaelan Connell, Cristiano Costa, Ben Csiernik, Stephanie Di Pelino, Junior Vitorino Fandim, Shireen Harbin, Dr Wilhelmina IJzelenberg, Carsten Bogh Juhl, Mariana Leite, Alanna MacDonald, Devin Manning, Daniele Sirineu Pereira, Dr Diego Roger‐Silva, Heather Shearer, Danielle Southerst, Maria N Wilson, Jessica Wong, Leslie Verville, Hainan Yu. We appreciate advice and guidance from our Collaborative Review Working Group members: Jan Hartvigsen, Chris Maher, Andrea Furlan, Toby Lasserson, Peter Tugwell, Maurits van Tulder, Amir Qaseem, Manuela Ferreira, Rachelle Buchbinder. Thank you to Victoria Pennick for helpful copy editing. The methods section is based on the standard Cochrane Musculoskeletal protocol template and the Cochrane Methods Support protocol template.
Appendices
Appendix 1. Preliminary MEDLINE Ovid search strategy
| MEDLINE All Ovid | |
| 1 | randomized controlled trial.pt. |
| 2 | controlled clinical trial.pt. |
| 3 | pragmatic clinical trial.pt. |
| 4 | random*.ti,ab. |
| 5 | placebo.ab,ti. |
| 6 | drug therapy.fs. |
| 7 | trial.ab,ti. |
| 8 | groups.ab,ti. |
| 9 | or/1‐8 |
| 10 | (animals not (humans and animals)).sh. |
| 11 | 9 not 10 |
| 12 | exp Back Pain/ |
| 13 | Intervertebral Disc Displacement/ |
| 14 | exp Sciatic Neuropathy/ |
| 15 | exp Spondylosis/ |
| 16 | (back ache* or backache* or back disorder* or back pain*).tw,kw,kf. |
| 17 | coccydynia.tw,kw,kf. |
| 18 | ((disc? or disk?) adj1 (degenerat* or displace* or hernia* or prolapse* or slipped)).tw,kw,kf. |
| 19 | dorsalgia.tw,kw,kf. |
| 20 | (lumb* adj4 pain).tw,kw,kf. |
| 21 | lumbago.tw,kw,kf. |
| 22 | (sciatic neuropathy or sciatica or ischialgia).tw,kw,kf. |
| 23 | (spondylosis or spondylolysis or spondylolisthesis).tw,kw,kf. |
| 24 | or/12‐23 |
| 25 | exp Exercise/ |
| 26 | exp Exercise Therapy/ |
| 27 | exp Exercise Movement Techniques/ |
| 28 | Physical Therapy Modalities/ |
| 29 | exp Recreation/ |
| 30 | Recreation Therapy/ |
| 31 | exp Physical Fitness/ |
| 32 | exercis*.tw,kw,kf. |
| 33 | (kinesiotherapy or recreation*).tw,kw,kf. |
| 34 | McKenzie.tw,kw,kf. |
| 35 | Alexander.tw,kw,kf. |
| 36 | William.tw,kw,kf. |
| 37 | Feldenkrais.tw,kw,kf. |
| 38 | (McGill adj5 (method or technique)).tw,kw,kf. |
| 39 | (training adj2 (strength* or physical or fitness or core or ergonomic* or musc* or spine or spinal or balance or stabil*)).tw,kw,kf. |
| 40 | ((core or musc*) adj2 (strengthen* or stabiliz* or stabilis* or stability or endurance or condition*)).tw,kw,kf. |
| 41 | functional restoration.tw,kw,kf. |
| 42 | pilates*.tw,kw,kf. |
| 43 | (yoga or hatha or ashtanga or bikram or iyengar or kripalu or kundalini or sivananda or vinyasa or raja or radja or bhakti or jnana or kriya or karma or yama or niyama or asana or pranayama or pratyahara or dharana or dhyana or samadhi or bandha or mudra or yin).tw,kw,kf. |
| 44 | aerobic*.tw,kw,kf. |
| 45 | (high intensity interval training or hiit).tw,kw,kf. |
| 46 | (walk* or run or running or jog or jogging or sport* or cycling or biking or swim* or dance or dancing or gymnastic* or boxing or kickboxing or stretch*).tw,kw,kf. |
| 47 | (aquacise or aquacize or aquasize or aquafit* or zumba or barre).tw,kw,kf. |
| 48 | (tai chi or tai ji or taiji or taijiquan or taijizhang).tw,kw,kf. |
| 49 | eldoa.tw,kw,kf. |
| 50 | (glad adj5 (hip? or knee? or osteoarthritis)).tw,kw,kf. |
| 51 | (otago adj5 (program* or balance or strength or training)).tw,kw,kf. |
| 52 | (bone fit or bonefit).tw,kw,kf. |
| 53 | walk tall.tw,kw,kf. |
| 54 | (dynamic neuromuscular stabili?ation or dns).tw,kw,kf. |
| 55 | active rehabilitation.tw,kw,kf. |
| 56 | or/25‐55 |
| 57 | Alexander Disease/ |
| 58 | Williams Syndrome/ |
| 59 | or/57‐58 |
| 60 | 56 not 59 |
| 61 | 11 and 24 and 60 |
Appendix 2. Search optimisation report. Exercise for low back pain
Literature search optimisation report (prepared by Leah Boulos, Maritime SPOR SUPPORT Unit, Nova Scotia, Canada)
Gold standard
The gold standard for this optimisation experiment was derived from studies included in the Exercise for Low Back Pain Cochrane review, as well as a list of potentially relevant studies not included in the review. The total number of studies in the gold standard was 305.
Database coverage
98% of the studies in the gold standard (n = 298) were found in CENTRAL, MEDLINE, or Embase. 1% of the studies (n = 4) were found in CINAHL, SPORTDiscus, or PEDro. There were no unique studies found in PsycINFO. 1% of the studies (n = 3) were not found in any of the seven databases searched for this review. This data supports searching only CENTRAL, MEDLINE, and Embase for this review.
Proposed revisions to exercise portion of search
Based on a keyword analysis of studies not retrieved by the MEDLINE search, the following changes are proposed.
| Original search | Revised search | ||
| 1 | exp Exercise/ | 1 | exp Exercise/ |
| 2 | exercis*.tw,kf. | 2 | exp Exercise Therapy/ |
| 3 | exp Exercise Therapy/ | 3 | exp Exercise Movement Techniques/ |
| 4 | exp Exercise Movement Techniques/ | 4 | Physical Therapy Modalities/ |
| 5 | exp Physical Therapy Modalities/ | 5 | exp Recreation/ |
| 6 | McKenzie.tw,kf. | 6 | Recreation Therapy/ |
| 7 | Alexander.tw,kf. | 7 | exp Physical Fitness/ |
| 8 | William.tw,kf. | 8 | exercis*.tw,kf. |
| 9 | Feldenkrais.tw,kf. | 9 | recreation*.tw,kf. |
| 10 | exp Yoga/ | 10 | McKenzie.tw,kf. |
| 11 | exp Recreation/ | 11 | Alexander.tw,kf. |
| 12 | exp Physical Fitness/ | 12 | William.tw,kf. |
| 13 | (yoga or pilates).tw,kf. | 13 | Feldenkrais.tw,kf. |
| 14 | (Tai Chi or Tai Ji or Taiji or Taijiquan).tw,kf. | 14 | (training adj2 (strength* or physical or fitness or core or ergonomic* or musc* or spine or spinal or balance or stabil*)).tw,kf. |
| 15 | or/1‐14 | 15 | (core adj2 (strengthen* or stabiliz* or stabilis* or stability)).tw,kf. |
| 16 | exp Alexander Disease/ | 16 | functional restoration.tw,kf. |
| 17 | exp Williams Syndrome/ | 17 | (pilates* or yoga).tw,kf. |
| 18 | 16 or 17 | 18 | aerobic*.tw,kf. |
| 19 | 15 not 18 | 19 | (walk* or run or running or jog or jogging or sport* or cycling or swim* or dance or dancing or gymnastic* or boxing or kickboxing).tw,kf. |
| 20 | (aquacise or aquacize or aquasize or aquafitness or zumba or barre).tw,kf. | ||
| 21 | (tai chi or tai ji or taiji or taijiquan).tw,kf. | ||
| 22 | active rehabilitation.tw,kf. | ||
| 23 | or/1‐22 | ||
| 24 | Alexander Disease/ | ||
| 25 | Williams Syndrome/ | ||
| 26 | or/24‐25 | ||
| 27 | 23 not 26 | ||
Search characteristics
Below is a comparison of characteristics for the original vs revised searches. The search characteristics in CENTRAL remain essentially unchanged. In MEDLINE, sensitivity increases from 95.8% to 97.7%, while the number needed to read (NNR) decreases from 18 to 17. In Embase, sensitivity increases from 90.2% to 95.7%, but the NNR increases from 25 to 35.
| Search Strategy | Comprehensiveness/ sensitivity/ recall ‐ a/(a+c) | Efficiency/ precision ‐ a/(a+b) | NNR ‐ 1/(a/(a+b)) | Specificity ‐ d/(b+d) | Accuracy ‐ (a+d)/(a+b+c+d) |
| CENTRAL | |||||
| Original | 95.5% | 4.6% | 22 | 80.9% | 81.1% |
| Revised | 95.2% | 4.3% | 23 | 79.9% | 80.1% |
| MEDLINE | |||||
| Original | 95.8% | 5.4% | 18 | 80.2% | 80.4% |
| Revised | 97.7% | 5.8% | 17 | 81.2% | 81.4% |
| Embase | |||||
| Original | 90.2% | 4.0% | 25 | 90.3% | 90.3% |
| Revised | 95.7% | 2.9% | 35 | 85.4% | 85.4% |
Records for screening
If a search update were run today (1 September 2021) for new results published in 2021 only, the results would be as follows:
| Original search | Revised search | ||
| CENTRAL | 275 | CENTRAL | 313 |
| MEDLINE | 232 | MEDLINE | 248 |
| Embase | 311 | Embase | 461 |
| CINAHL | 262 | ||
| PsycINFO | 13 | ||
| SPORTDiscus | 55 | ||
| PEDro | 25 | ||
| Total results | 1173 | Total results | 1022 |
| Duplicates removed | 400 | Duplicates removed | 289 |
| Total to screen | 773 | Total to screen | 733 |
Conclusions
Running the revised version of the exercise portion of the search in fewer databases would result in an increase in sensitivity while slightly reducing the number of records to screen.
Appendix 3. Guidance documentation for all study characteristics extracted
|
Data extraction guidelines. Exercise therapy for chronic low back pain (Collaborative Review) Please note that this is intended to be a living document, and is updated regularly to provide further clarification for data extraction items. The most recent version of this document is to be linked at the top of the data extraction form within DistillerSR for ease of reference for all team members. (Developed by the BACK Program team, Dalhousie University. Halifax, NS, Canada. Please cite as: Hayden JA, Ogilvie R, Singh S, Ellis J. Data extraction guidelines: Exercise therapy for chronic low back pain (Collaborative Review). 2022; version May 23, 2022) | ||
| **This symbol/colour shading indicates branching fields (i.e. will only be visible if the appropriate response to a previous question was selected) | ||
Please be consistent with data entry convention for free text boxes:
| ||
| Study Information | ||
| Study ID (First author last name, year of publication ‐ e.g. Smith 2007) | Format: Last name YYYY | |
| Corresponding author last name | Record the corresponding author’s last name (even if they are the first author). | |
| Corresponding author email address | Record the corresponding author’s email address. | |
| Trial registration number reported (if yes, record) | Record the trial registration number, if yes. Format: NCT01343927 | |
| Published protocol reported (if yes, record citation) | Record the protocol citation, if yes. Note: if you discover that a study has supplements or another materials necessary for full extraction, please email research coordinator. | |
| Research ethics board review reported (if yes, record name of the Review Board and REB#, if available) | Record if the trial reported receiving ethical approval. Format: Boston University Institutional Review Board This study had approval from the SPECIFIC Research Ethics Board (e.g. Dalhousie University Research Ethics Board): Yes This study had institutional approval from SPECIFIC university (e.g. approved by Dalhousie University) and followed Helsinki Declaration or followed by informed consent (word ethics is not stated): Yes This study had ethical approval from our local institution (specific institution not provided): Yes |
|
| Conflict of interest statement reported (if yes, record conflict statement or if study reports that there were no conflicts of interest, record as 'None to declare’) | Copy and paste the conflict statement in the manuscript. If the authors declare no conflicts ‐ use this language in the Distiller text box: None to declare Format: ‘Dr. Saper reports grants from the National Center for Complementary and Integrative Health of the National Institutes of Health during the conduct of the study.’ Format: None to declare |
|
| Funding source for trial reported (if yes, record the funding source or if study was not funded, record as 'No funding received') | Record the name of the funding source(s), separating funding sources by a semicolon. If the trial reports that they did not receive funding, use statement: No funding received Format: National Center for Complementary and Integrative Health of the National Institutes of Health; Centre for Yoga Therapy Format: No funding received |
|
| Study participant recruitment dates reported (format: December 2015 to December 2017) | Record all dates provided in the publication.
Format: December 2009 to March 2010 OR 2009 to 2010 ‐ depending on detail given in publication |
|
| Number of participants randomised (all groups combined) | Record number of participants randomised into the trial. | |
| Are the majority of study participants older adults (age ≥ 60 years) OR is there a subgroup of participants of older adults aged ≥ 60 years presented? | Must be an older adult population (majority of participants is ≥ 60 years). Trial can indicate inclusion criteria greater than 60 years or report a subgroup of participants years and older. If the study population is mixed age, then consider the mean and standard deviation to decide if > 75% is older than 60 years [e.g. mean age – (0.675 x SD) > 60 y, assuming a normal distribution). Indicate unclear if not clear from the full text publication. | |
| Country/countries of study conduct | Record the full name of the country or countries of conduct, separating countries by a semicolon. If the country of conduct is not reported, record the country of the first author’s institution. Format: United States; Canada; Denmark (full name of country) |
|
| Population low back pain (LBP) duration | Record the LBP duration in the study population.
|
|
| Inclusion criteria for pain duration (include units ‐ days/weeks/months) | Record the specific inclusion criteria that the study used with reference to symptom duration of LBP (e.g. ≥ 6 weeks, ≥ 2 months, etc.). Format: 3 months; ≥ 12 weeks |
|
| Pain duration (include units ‐ days/weeks/months) | Record the duration of LBP episode among all the participants randomised in the study. Prioritise extracting mean duration of population if it is provided. If not, extract % when possible. Extract by study group if study population duration is not available (example, Ex 1 – 12 weeks Ex 2 – 13 weeks, Comp 1 – 12 weeks) | |
| Presence of leg pain and/or lower limb neurological symptoms (e.g. weakness, sensory deficits) | Record the proportion of the study population that had leg pain and/or lower limb neurologic symptoms.
|
|
| Inclusion of any participants with specific‐cause LBP (if yes, what %) | Record the % of participants included in the trial with specific‐cause low back pain (must be less than 25% of participants to be eligible for inclusion, unless subgroups are presented) Format: 12% |
|
| Population setting or source | Record the recruitment location for the study population.
|
|
| Include or exclude study | Record if the study should be included or excluded. Stop extracting if you choose to exclude. | |
| **Reason for exclusion | Record the first reason for exclusion for this study (following the order presented).
Complete any mandatory questions with Not Reported or 9999 and click submit. |
|
| Study information notes | Record any extra or unique information about the study (e.g. special study selection criteria, such as study was limited to military personnel; study was limited to people with average pain intensity in the previous week of 4 or greater on an 11‐point (0 to 10) numerical rating scale). | |
| Network review primary outcomes reported | Record all the network review primary outcomes that the study measured and reported. (NOTES: primary outcome refers to our review’s primary outcomes NOT the trial’s primary outcomes. If a study appears to report other outcomes in a separate publication, not yet in hand, record the citation (if available) and outcomes in the very last text box in this form). Note ALL scales reported for each outcome below. For example, if both the RMDQ and ODI are reported for the functional limitations outcome, record both scales in the text box provided (but extract data only for RMDQ). Check the box for Adverse events if adverse events are mentioned in the publication (please CNTL F “adverse”).
|
|
| Describe how adverse events are reported | Record information about how the trial collected information related to adverse events. If adverse events outcome data are not presented by study group, please include results here. If available, describe whether the adverse events were considered to be related to the intervention and how this was determined within each trial (e.g. by trialists or by an independent monitoring board) | |
| Were adverse events measured systematically for all study participants, or only recorded ‘as reported’ by participants? | Record if adverse events were:
|
|
| Network review secondary outcomes reported (indicate specific measure/scale) | Record all the network review secondary outcomes that were measured and reported in the study by checking all that apply, then listing all other available outcomes in the 'other' checkbox. For the listed outcomes, indicate the scale/measure used to assess each secondary outcome in the available text boxes. (NOTES: secondary outcome refers to our review’s secondary outcomes NOT the trial’s secondary outcomes. If a study appears to report other outcomes in a separate publication, not yet in hand, record the citation (if available) and outcomes in the very last text box in this form). | |
|
|
|
| Pain/Function/Health‐related QoL/Depression measures/scales | Record the measure/scale used to assess the outcome of interest. If there are multiple measures/scales for an outcome, prioritise the following measures in the following order, when available:
|
|
| Describe Pain/Functional limitations/Health‐related QoL/Depression measure/scale | Copy and paste any additional information about the measure/scale provided in the publication. | |
| Considering Pain/Functional limitations/Health‐related QoL/Depression measure/scale direction, which is true? | Record which direction the measure/scale should be interpreted.
|
|
| Minimum value of Pain/Functional limitations/Health‐related QoL/Depression measure/scale | Record the minimum value of the measure/scale used to assess the outcome of interest. Format: 0 |
|
| Maximum value of Pain/Functional limitations/Health‐related QoL/Depression measure/scale | Record the maximum value of the measure/scale used to assess the outcome of interest. Format: 100 |
|
| All available follow‐ups | Record all available follow‐up periods. Note: When extracting outcome data later on in the form, only extract immediate and very long‐term follow‐up data if no data are available for short‐term or long‐term follow‐up. To categorise a follow‐up period, follow the two steps below: 1. Put it in the category within its time frame: 6 to 12, 13 to 47, 48+ wks 2. If a single study has two measurement periods falling within one follow‐up period, the one that gets selected is the one closest to 3, 6, 12 mo. If there is more than 2 weeks between randomisation and the start of treatment, please note this in the notes section.
|
|
| **Immediate‐term follow‐up (< 6 weeks): Indicate unit of measure reported and exact # of weeks/months/other | Record the exact length of the immediate‐term follow‐up period. First select the unit of time used to measure, then state the exact length of time. | |
| **Short‐term follow‐up (closest to 3 months): indicate unit of measure reported and exact # of weeks/months/other | Record the exact length of the short‐term follow‐up period. First select the unit of time used to measure, then state the exact length of time. | |
| **Moderate‐term follow‐up (closest to 6 months): indicate unit of measure reported and exact # of weeks/months/other | Record the exact length of the moderate‐term follow‐up period. First select the unit of time used to measure, then state the exact length of time. | |
| **Long‐term follow‐up (closest to 12 months): indicate unit of measure reported and exact # of weeks/months/other | Record the exact length of the long‐term follow‐up period. First select the unit of time used to measure, then state the exact length of time. | |
| Outcomes and follow‐up notes | Record any unusual or unique details about the outcomes and their measurement, as well as the follow‐up time periods. | |
| Study Design | ||
| Number of exercise groups | Record the number of exercise groups described in the publication. This will generate the appropriate data extraction fields below. | |
| Number of non‐exercise comparison groups | Record the number of non‐exercise comparison groups described in the publication. This will generate the appropriate data extraction fields below. | |
| IMPORTANT: Please remember to extract the exercise groups in the order that they appear in the publication. Exercise Group 1 should be the first exercise group, Exercise Group 2 should be the second exercise group described, etc. Comparison Group 1 should be the first non‐exercise comparison group described. | ||
| Exercise Groups | ||
| Treatment description | Record a thorough description of the treatment delivered for this treatment group. | |
| All exercise types delivered to this group (check all that apply) | Record all exercise types delivered to the treatment group, regardless of the proportion of the treatment the exercise represented. | |
| Dominant exercise type(s) delivered to this group (select up to two types, or mixed exercise if 3 or more types are equally dominant) | Record the dominant exercise type(s) delivered to this group (select up to two types, or mixed exercise if 3 or more types are equally dominant).
|
|
| Was this exercise treatment sufficiently robust that it would be expected to be effective by both the research team, provider and participants? | Record if the treatment delivered in this group was a robust and plausible intervention designed for maximum outcome improvement, such that the research team, provider and participants would expect it to be effective. This is the data extractor’s judgement. | |
| Does this group include a graded‐activity approach to the treatment delivery? | Graded activity = graded activity includes baseline assessment to identified problematic activities and exercise progressed using a time‐contingent manner, regardless of pain, with use of quotas and pacing. Graded activity includes cognitive and/or behavioural principles such as operant conditioning with the use of strategies of positive reinforcement, and reassurance. The reporting of a gradual increase of exercise dosage doesn’t by itself characterise GA. | |
| Is the treatment delivered to this group community‐based exercise? | Community‐based exercise = community‐based exercises are exercise treatments delivered by a non‐healthcare professional. A non‐healthcare professional does not have advanced healthcare training and education and is not a member of a licensed, regulated health profession (e.g. community‐health workers or personal trainers are NOT healthcare professionals). Exercise treatments intended to be done alone (e.g. home exercise program) that are delivered by a healthcare professional, in‐person or through telehealth are not community‐based. | |
| Exercise specificity | Record the correct exercise specificity.
|
|
| Program design individualisation | Record the level of individualisation of the treatment.
|
|
| Program delivery site | Record the delivery site of the treatment.
|
|
| Program delivery mode | Record the delivery mode of the treatment.
|
|
| Home exercise component | Record if the treatment included a home exercise component | |
| Indicate all co‐interventions provided to participants in this exercise group | Record all co‐interventions delivered to this group or indicate if no co‐interventions were delivered to the group. | |
| ||
| **All participants in group received identical co‐intervention(s) | Record if all participants in the group received identical co‐intervention | |
| **YOGA QUESTION SET | See end of document for further information about yoga‐specific questions | |
|
Dose and duration table For example: an exercise treatment was delivered in 30‐minute sessions (i.e. dose (hours per session)) with 6 sessions delivered over the course of 2 weeks (i.e. program duration (weeks)). If the exercise treatment included a clinic or community component and recommended home exercises, record information about the main exercise sessions below; record information about any home exercises in the dosage/adherence comment box below. | ||
| Dosage (minutes per session) | Record how long each session of exercise treatment delivery lasted, in minutes. Format: 60 |
|
| Number of sessions | Record the total number of sessions that each study participant was asked to attend in the treatment program. Format: 12 |
|
| Program duration (weeks) | Record the total length of time over which the treatment was delivered. Format: 12 |
|
| Therapist delivering exercise treatment | Record the professional designation of the therapist that delivered the treatment to this group.
|
|
| Materials/equipment required for exercise program (if yes, what?) | Record if special materials or equipment were required for the treatment delivered to this group. It is not necessary that all participants used all required equipment. If yes, indicate what materials were required (e.g. yoga block, Pilates reformer, stability ball, hand weights) | |
| Cost to participants (if yes, how much?) | Record if there was a cost to participants for the treatment delivered to this group. If yes, indicate the total cost of all sessions. Format: $250 |
|
| % adherence/completion | Record the extent to which participants followed the study requirements. If adequate adherence is reported in the study, report this value. e.g. the mean number of main sessions attended was 8 out of 10 = 80% If adherence is not specifically reported, then this should be estimated from information about participation based on the dominant site/approach of delivery following best estimates:
For example, if a program includes a predominantly clinic‐based supervised program with a home component, estimate based on “directly supervised by provider”. Format: 100% |
|
| Dosage and adherence notes | If necessary for clarification, describe the exercise treatment component(s) that the dose and duration above refers to; if there are additional components (e.g. home exercises not captured in a study‐defined adherence measure) describe this here and note any available adherence features of the group. | |
| Participant Description Table | ||
| Number of participants randomised to this group | Record the number of participants originally randomised into this group. Format: 100 |
|
| Sex (% male) | Record the percentage of the population that is male. Format: 54% |
|
| Age (mean or median) | Record the mean or median age of group participants. If age is only described for the whole study population and not by study group, then record it in Exercise Group 1 and make a note in the notes section of this table of that fact. | |
| Group description notes | Record any unusual or unique details about the basic characteristics of this treatment group population. | |
|
Outcome Measurement Tables All follow‐up time points will be displayed in the outcome tables regardless of what follow‐up time points are presented in the trial. Please leave cells empty (blank) in the column if the follow‐up time point is not presented. Prioritised the extraction of data in order above, as available: group follow‐up data based on intention‐to‐treat analysis (if unadjusted and adjusted are available, extract the primary reported), within group change scores, between group change scores, effect size. | ||
| Please indicate the measure of central tendency that is reported for the exercise group(s) in this study (choose mean, if available): | Record what type of statistic was used to measure the central tendency of pain outcome. Report mean, if available. | |
| Please indicate the measure of spread that is reported for the exercise group(s) in this study (choose SD, 95% CI if available): | Record what type of statistic was used to measure the spread of pain outcome. Report standard deviation if available, or 95% confidence interval, if standard deviation is not reported. Note: interquartile range is the same as (Q1, Q3). | |
| What type of data are you extracting for this outcome? |
Record the type of data you are extracting for the outcome. Prioritised the extraction of data in order below, as available: group follow‐up data based on intention‐to‐treat analysis (if unadjusted and adjusted are available, extract the primary reported), within group change scores, between group change scores, effect size.
|
|
| Outcome notes | Please use this notes box, provided after each outcome table under Exercise Group 1, to note any important information about the outcome data reported or extracted. | |
| All outcome tables | Extract data as reported in the publication, keeping the same number of decimal places. Do not round. Please leave cells empty (blank) in the column if the FU time point is not assessed in the trial. Only use 9999 if the follow‐up time point is assessed in the trial, but data are not presented for the outcome. | |
| Describe minor adverse events for this group, including n (%), when available | Record the n and/or (%) of group that reported minor adverse events. If adverse events outcome data are not presented by study group, please record all results in exercise group 1. Format if both n and % reported: 14 (80%) experienced headache, minor aches |
|
| Describe major adverse events for this group, including n (%), when available | Record the n and/or (%) of group that reported major adverse events. If adverse events outcome data are not presented by study group, please record all results in exercise group 1. Format if both n and % reported: 1 (3%) experienced cardiac arrest |
|
| Comparison Groups | ||
| Comparison group description | Record the treatment this group received (if any), including intensity/time/number of sessions. | |
| Comparison type | Record the comparison type of this group.
|
|
| **Other conservative treatment type (check all that apply) | Record the ‘other conservative treatment’ type delivered to this group. See definitions above for co‐interventions. | |
| **This treatment group was a main treatment of interest of the trial | Record if this treatment was a main treatment of interest of the trial. | |
| Number of participants randomised to this group | Record the number of participants originally randomised into this group. Format: 100 |
|
| Sex (% male) | Record the percentage of the population that is male. Format: 52% |
|
| Age (mean or median) | Record the mean or median age of group participants. If age is only described for the whole study population and not by study group, then record it in Exercise Group 1 and make a note in the notes section of this table of that fact. | |
| **Adverse events reported for group (if minor or major, describe)? | Record % of group that reported minor or major adverse events. If adverse events outcome data are not presented by study group, please record all results in exercise group 1. Format: 60% |
|
| Group description notes | Record any unusual or unique details about the basic characteristics of the comparison group population. This notes box appears only for Comparison Group 1; if more than 2 comparison groups are included in the study, please make notes about all comparison groups here. | |
| Additional study notes | Record the citations for any linked publications referenced in this trial publication, additional supplements, etc. | |
| YOGA QUESTION SET | This set of questions will appear if yoga is selected as a dominant exercise type in any exercise group. | |
| **Characterisation of the type or style of yoga (check all that apply) | Record the named type of yoga delivered to participants. Check all that apply.
|
|
|
**Components of the yoga treatment (check all that apply) |
Record all components of the yoga treatment delivered to participants. Check all that apply.
|
|
| **Are specific poses listed or pictured? (check all that apply) | Record the information provided about specific posed delivered to participants. Check all that apply.
|
|
| **How was the yoga treatment designed? | Copy and paste the quote from the report regarding how the treatment was designed. | |
| **Does the report describe monitoring for treatment fidelity? | Record if the report described monitoring for treatment fidelity.
|
|
| **What does the report state about flexibility of the treatment delivery? | Record the level of flexibility of delivery of the yoga treatment delivered to participants.
|
|
| **What does the report state about the background of the teachers? | Record the level of training of the yoga teachers delivering the yoga treatment.
|
|
Appendix 4. Data cleaning and validation guidelines
Data Extraction Guidelines: Exercise Therapy for Chronic Low Back Pain (Collaborative Review)
(Developed by the BACK Program team, Dalhousie University. Halifax, NS, Canada. Please cite as: Kashif S, Hayden JA. Data cleaning and validation guidelines: Exercise therapy for chronic low back pain (Collaborative Review). 2022; version June 28, 2022
There are always some sources of errors in systematic data collection process irrespective of the procedures and error preventing measures which can cause inconsistencies in data. These errors include but are not limited to measurement errors, entry errors, context errors, and processing errors. These inconsistencies can inflict erroneous data analysis. The major concern before data analysis should be to ensure the accuracy and consistency of the data, which necessitate a systematic and planned way for identifying and treating errors (ACAPS 2016; Broeck 2005).
Limited guidance is available in the literature for the methodologies and standards for optimum data cleaning for systematic review. We present our proposed approach for this collaborative systematic review below.
Data generation process for systematic review
The robust methods, including clear data collection forms and guidance documentation, will be used for systematic review data collection. Also, all reviewers conducting data extraction will complete formal training and a calibration exercise. Distiller SR systematic review management software will be used to track data extraction and facilitate consensus. The process for data extraction will include one review team member extracting study data, including population, intervention(s), comparison(s), outcome information (measure and timing), and completing the Cochrane RoB 1 tool for each study, onto pre‐tested standardised DistillerSR forms (Distiller SR; Higgins 2011). A second reviewers will conduct a careful quality check of the full data extraction, with discussion and consensus decisions for any discrepancies. Data collected in past updates of this review will be checked.
Sources of anomalies in systematic review data
The common sources of erroneous data points and anomalies in systematic review data that we will consider are:
Errors in data extraction (e.g. measurement errors, syntax errors, recording error);
Misunderstanding by reviewers;
Misreporting by trialists;
Integrity issues (e.g. data fabrication or falsification).
Objectives of data cleaning
The main objectives for using this data cleaning guideline are to:
Discover and correct data extraction errors;
Detect and note any study data extracted correctly but flagged for potential integrity concerns (for sensitivity analyses);
Construct a standardised data format to be used in different research projects;
Structure more understandable and reproducible data in terms of transparency and accountability.
Standards for systematic review quality data
Quality data are crucial for a reliable analysis, and hence, for concerned inferences. The decisions based on poor data may be misleading, and hence, have less confidence. The broad standards of a quality data include: accessibility, accuracy, comparability, completeness, consistency, coherence, credibility, reliability, flexibility, plausibility, relevance, usefulness, timeliness, uniqueness, and validity (ACAPS 2016; Britan 2009). For this systematic review database, we will focus on the five quality standards of validity, accuracy, completeness, uniformity, and integrity.
Data validity
Data validity refers to the degree to which data points essentially represent what is anticipated to be measured (Britan 2009), which implies that the data should conform to defined restrictions for different variables and values in the data. For systematic review data, we will focus on the restrictions for variable types, variable range, mandatory variables restriction, unique variable restriction, set membership restriction, regular expression patterns restriction, cross‐field validation restriction and randomness restriction (Table below).
Data accuracy
Data accuracy implies the extent to which the data points are expected to be close to the true prespecified values. Applying all data validity constraints does not mean that all valid values are accurate and also precise. So, further checks, such as cross‐tab evaluation, association or correlation should be applied for data accuracy.
Data completeness
Data completeness defines the level to which all required data are known and complete. There are various reasons for missing data, and we will improve and lessen the problem by contacting the original source (trial authors, using REDCap survey procedure), if possible (Harris 2019).
Data uniformity
Data uniformity represents the level to which the data are quantified, using the same defined unit of measure for different variables. And so, to obtain data uniformity, some data will be converted to a single measure unit for different variables. For example, outcome data converted to a common scale as appropriate, exercise frequency expressed as common time measure.
Data integrity
Data integrity indicates the correctness of the collected data without intended manipulation. This implies that all data from all studies should be free from data exploitation for any reason. Data integrity problem may arise during transcription, which can be checked and corrected; or more seriously, due to deliberate manipulation, which we will examine using randomness test or graphical methods and addressed accordingly.
Table: Data validity restrictions
| Restrictions type | Description |
|
Variable type restriction |
The variables in the data should be of a particular data type, e.g. character, numeric, date, etc. |
|
Range restriction |
Values in a particular variable (numeric) should fall within a certain specified range. |
| Mandatory restriction | Certain variables in data cannot be empty. |
| Unique restriction | A variable, or a combination of variable, must be unique across data. |
| Set‐membership restriction | The values of a categorical variable come from a set of specified values. |
| Regular expression patterns restriction | The character/text variables should be in a certain defined pattern. |
| Cross‐field validation restriction | The different variables in the data set should meet the certain conditions that span across in other variables. |
|
Randomness restriction |
The certain numeric variable values or a group of values should follow the random pattern for accuracy and unbiasedness. |
Former data cleaning measures
After data extraction and importing, the first measure will be to save the original data file with a specified format, and make a copy of the original data for further process. The second measure will be to consider the two main pieces of information as:
Unit of observation of the data;
Unique and fully identified ID variables of the data.
Unit of observation of the data will involve two check points.
What does each row represent in the data?
Which variable in the data will be the main unique ID identifying each row?
After this, the first check will be identifying duplicated IDs in the data; correction will be made after consensus.
Steps in data cleaning
Data cleaning is defined as the recurrent cycle of various stages of consist of screening, diagnosing, and editing with the procedures and methods to deal with data problems and errors (Broeck 2005).
We will use the R package for all data scrubbing measures (Jonge 2013). For this systematic review database, data cleaning will be an iterative process of a sequence of five phases, to produce high‐quality data while considering data quality standards at different phases.
1. Data inspection
The data inspection will be the detection of structural errors by diagnosing the unexpected, incorrect, and inconsistent data points. The data inspection will be a time‐consuming procedure, which involves many methods for exploring the raw data for error detection, and we will focus and define a priori the most important variables for checking (and types of checks for these select variables) to be feasible. The Exploratory Data Analysis is the most used method for data inspections and rectification of data errors. This involves data profiling and data visualisation to describe the possible structure and values in the data using statistical properties of the data (Dasu 2003; Hellerstein 2008; Tukey 1977).
Data profiling
The data reporting/profiling using summary/descriptive statistics will be beneficial to overview the structure of data points. Using statistical methods, such as measure of central tendencies and dispersions to investigate data will underline the unexpected and thus erroneous data points. Similarly, data tabulation will be applied to confirm the unique values and logical consistency between different variables. For example, these approaches will be useful to:
Check types of recorded variables as character or numeric;
Test the range of data variables lies within the specific values;
Identify the unique values in different variables;
Conform given standard of certain variables;
Confirm different logical conditions between values of different variables in the data, and hence, to flag inconsistent variables and values;
Report the overall missing data points or in certain variables.
Data visualisation
Using different plots for data visualisation will be effective to recognise unique values, distributions, patterns, and irregular data points in variables. Visualising the data using bar and box plots and distribution plots will help to identify inconsistent data points. For example, data visualisation will be worthwhile to:
Identify the form of the distribution for different variables;
Discover the proportion of unique values in different variable;
Conform the certain pattern in different variable;
Detect outlier data points;
Distinguish influential data points;
Discern the distribution of missing data points in different variables to create flag variables for potential description.
2. Data verification
The study data points flagged in the above step will be double‐checked against the publication report for errors in data extraction, and misunderstanding by reviewers in the data verification phase.
Influential data points confirmed to agree with the study report, but believed to be erroneous, will be checked with the trial's corresponding author to verify the study data. After data verification, we will classify suspected data points as:
Erroneous;
True extreme;
Idiopathic.
3. Data correction
Data correction will involve different techniques and methods based on the identification and irregularities of the data points. In data correction, we will rectify data points if they are erroneous, and consider imputation (with sensitivity analysis) if idiopathic. We will make the decision for outlier/true extremes after investigating their effect on results (sensitivity analysis) and consensus.
3a. Methods for correcting erroneous data points
The erroneous data points will be fixed after identification of their types.
Irrelevant data
Irrelevant variables and data points will be those that will not actually be needed, and created during the data recording and extraction process. We will check data by column for irrelevant variables, and by row for irrelevant observations/levels, and will drop an unimportant piece of information after we are sure.
Duplicate elimination
As data will be combined from different sources, there will be odds of having duplicate data points. A common indicator will be when two studies have the same reference ID or study ID. And therefore, we will simply remove them after confirmation.
Type conversion and creation of new variables
The variables should be in the specified types and formats. This will be checked quickly by looking over the variable types in each column summary, and change them, as required.
Similarly, new labelled variables as categorical variables will be generated from numerical variables, and new numerical variables will be created from character variables, if needed.
The variables or values that can’t be transformed to a specified type will be converted to NA value (or any specified format), with a warning inserted to indicate which variable/value is incorrect and must be fixed with consultation.
Syntax errors
The syntax errors will be fixed.
The additional white spaces at the beginning, end or in the middle of character/string variables will be deleted.
The character/strings variables will be padded with spaces or other characters to a certain width.
The typos in the character/strings will be fixed, as these variables will be entered in many different ways and can have mistakes. We will replace all values in such variables with one unique value.
Special considerations will be given to values, such as 0, Not applicable, NA, None, N/A, Null, 9999, or 7777, as they will represent missing values. These values will be changed to one unique value for consistency.
Consistency errors
The consistency errors between two variables identified in data inspection will be again confirmed and addressed accordingly.
Variable standardisation
All numerical and character variables will be changed to a standard format for data uniformity after all possible syntax errors are corrected.
For character/string variables, all values should be in lower or upper case.
For numerical variables, all values should be in a specified measurement unit and decimal points, as required.
Missing values
The identification of missing values as truly missing from studies, or missing in data extraction, will be important, and flagging will be used to distinct both. Also, to differentiate missing values from 'default' and 'unknown' values will be important, and flag variables will be used. The variable observations with missing values in systematic review data set will not be dropped. The variables with missing values may be missing, and will be reported as missing or imputed.
In a systematic review data set, imputation will be performed only for specific variables, and will be applied only for outcome measures of variance (SD). If mean outcome are available, but variance measure are missing or implausible, we will impute this SD. The previous method of imputation (which seems reasonable again) was to apply a simple imputation using the average SD from similar trials as an estimate of the missing value, and use it for imputation. 'Similar trials' will be based on two to three study characteristics (e.g. population source, outcome measure, treatment group) in the studies.
The flagged variables will be used to keep the missing data points information in the data set.
Scaling/normalisation
We will scale/normalise our outcome measures data based on calculated multiplicative factors for different outcomes, to minimise skewness for having different measurement units.
3b. Methods for correcting true extreme/outliers data points
The values of the variables that significantly differ from all other values, will be known as outliers. The outliers might be influential data points if they have an impact on model coefficients, and will deviate the model from where most of the data points lie. We will keep these values, unless there is a good reason to remove these values after investigation.
3c. Methods for correcting idiopathic data points
Idiopathic data points will be those data points that are still suspect, without any clarification found. We will impute such data points with sensitivity analysis, to determine whether to use the original data points or imputed values for further analysis. The similar, stated imputation method will be used for such data points.
4. Data re‐verification
After data cleaning and imputation, we will re‐inspect the data to validate accuracy, and for the set rules and constraints. Some manual corrections and adjustments will be made, if not possible otherwise.
5. Data reporting
We will prepare a report about the changes made to clean the data, as data reporting will be as considerable as data cleaning. Each corrected data file will be saved with a suffix with the date last modified, in the following format, to track the changes over time (i.e. Month. Date. Year. Version; for example, data.Apr.12.2022.v1.csv). Similarly, each code file will be saved with the same format and name to match the changes made in the data, and for reproduction. We will also construct a logbook with the same name and format to track the changes made in the data for each variable or values, with the causes of errors/irregularities that should be prevented in future projects and data extraction.
The intermediate files created during data cleaning will be saved separately, and named in a specified format; their links will be stored in the logbook for tracking and reproduction.
Database created versions
The procedures for data cleaning will include saving the study database at three stages:
Original data collected by reviewers;
Corrected (errors) after data cleaning;
Corrected and imputed, as appropriate.
These dataset versions will allow assessment of the impact of data cleaning on the overall results in systematic reviews.
Contributions of authors
All authors contributed to the drafting and revision of the protocol.
Sources of support
Internal sources
No sources of support provided
External sources
-
Canadian Institutes of Health Research (CIHR), Canada
CIHR provided funding for this project (Project Grant Competition, PJT‐173478, NPI Hayden)
Declarations of interest
JH: Canadian Institutes of Health Research (Grant / Contract)
SK: Canadian Institutes of Health Research (Grant / Contract)
RO: Canadian Institutes of Health Research (Grant / Contract)
SS: none
LB: none
SAS: none
LSW: none
FJM: none
BTS: none
TPY: none
AZ: none
KB: none
LA: none
GB: none
CC: none
New
References
Additional references
ACAPS 2016
- ACAPS. Data cleaning: dealing with messy data (technical brief). www.acaps.org/sites/acaps/files/resources/files/acaps_technical_brief_data_cleaning_april_2016_0.pdf 2016:1-17.
Alonso‐García 2020
- Alonso-García M, Sarría-Santamera A. The economic and social burden of low back pain in Spain: a national assessment of the economic and social impact of low back pain in Spain. Spine 2020;45(16):E1026-32. [DOI] [PubMed] [Google Scholar]
Beck 1987
- Beck AT, Steer RA, Brown GK. Beck Depression Inventory. New York: Harcourt Brace Jovanovich , 1987. [Google Scholar]
Beck 1988
- Beck AT, Steer RA, Carbin MG. Psychometric properties of the Beck Depression Inventory: twenty-five years of evaluation. Clinical Psychology Review 1988;8(1):77-100. [Google Scholar]
Bergner 1976
- Bergner M, Bobbitt RA, Pollard WE, Martin DP, Gilson BS. The sickness impact profile: validation of a health status measure. Medical Care 1976:57-67. [DOI] [PubMed] [Google Scholar]
Boulos 2022
Brignardello‐Petersen 2017
- Brignardello-Petersen R, Bonner A, Alexander PE, Siemieniuk RA, Furukawa TA, Rochwerg B, et al, GRADE Working Group. Advances in the GRADE approach to rate the certainty in estimates from a network meta-analysis. Journal of Clinical Epidemiology 2017;93:36-44. [DOI] [PubMed] [Google Scholar]
Britan 2009
- Britan G. Performance monitoring & evaluation tips data quality standards. USAIDS 2009;12:1-8.
Broeck 2005
- Broeck JV, Cunningham SA, Eeckels R, Herbst K . Data cleaning: detecting, diagnosing, and editing data abnormalities. PLoS Medicine 2005;2(10):e267. [DOI] [PMC free article] [PubMed] [Google Scholar]
Burckhardt 2003
- Burckhardt CS, Anderson KL. The Quality of Life Scale (QOLS): reliability, validity, and utilization. Health and Quality of Life Outcomes 2003;1(1):1-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
Caspersen 1985
- Caspersen CJ, Powell KE, Christenson GM. Physical activity, exercise, and physical fitness: definitions and distinctions for health-related research. Public Health Reports 1985;100(2):126. [PMC free article] [PubMed] [Google Scholar]
Chaimani 2015
- Chaimani A, Salanti G. Visualizing assumptions and results in network meta-analysis: the network graphs package. The Stata Journal 2015;15(4):905-50. [Google Scholar]
Chen 2022
- Chen S, Chen M, Wu X, Lin S, Tao C, Cao H, et al. Global, regional and national burden of low back pain 1990–2019: a systematic analysis of the Global Burden of Disease study 2019. Journal of Orthopaedic Translation 2022;32:49-58. [DOI] [PMC free article] [PubMed] [Google Scholar]
Chiarotto 2016
- Chiarotto A, Maxwell LJ, Terwee CB, Wells GA, Tugwell P, Ostelo RW. Roland-Morris Disability Questionnaire and Oswestry Disability Index: which has better measurement properties for measuring physical functioning in nonspecific low back pain? Systematic review and meta-analysis. Physical Therapy 2016;96(10):1620-37. [DOI] [PubMed] [Google Scholar]
Chiarotto 2018
- Chiarotto A, Boers M, Deyo RA, Buchbinder R, Corbin TP, Costa LOP, et al. Core outcome measurement instruments for clinical trials in nonspecific low back pain. Pain 2018;159(3):481-95. [DOI] [PMC free article] [PubMed] [Google Scholar]
Chiarotto 2019
- Chiarotto A, Maxwell LJ, Ostelo RW, Boers M, Tugwell P, Terwee CB. Measurement properties of Visual Analogue Scale, Numeric Rating Scale, and Pain Severity Subscale of the Brief Pain Inventory in patients with low back pain: a systematic review. Journal of Pain 2019;20(3):245-63. [DOI] [PubMed] [Google Scholar]
Chou 2010
- Chou R, Shekelle P. Will this patient develop persistent disabling low back pain? JAMA 2010;303(13):1295-302. [DOI] [PubMed] [Google Scholar]
Costa 2009
- Costa LC, Maher CG, McAuley JH, Hancock MJ, Herbert RD, Refshauge KM, et al. Prognosis for patients with chronic low back pain: inception cohort study. BMJ 2009;339:1-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
Costa 2019
- Costa N, Ferreira ML, Setchell J, Makovey J, Dekroo T, Downie A, et al. A definition of "flare" in low back pain: a multiphase process involving perspectives of individuals with low back pain and expert consensus. Journal of Pain 2019;20(11):1267-75. [DOI] [PubMed] [Google Scholar]
Dagenais 2008
- Dagenais S, Caro J, Haldeman S. A systematic review of low back pain cost of illness studies in the United States and internationally. The Spine Journal 2008;8(1):8-20. [DOI] [PubMed] [Google Scholar]
Dasu 2003
- Dasu T, Johnson T. Exploratory data mining and data cleaning. John Wiley & Sons, Inc., 2003. [Google Scholar]
Davidson 1986
- Davidson J, Turnbull CD, Strickland R, Miller R, Graves K. The Montgomery‐Åsberg Depression Scale: reliability and validity. Acta Psychiatrica Scandinavica 1986;73(5):544-8. [DOI] [PubMed] [Google Scholar]
Deeks 2022
- Deeks JJ, Higgins JP, Altman DG. Chapter 10: Analysing data and undertaking meta-analyses. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.3 (updated February 2022). Cochrane, 2022. Available from www.training.cochrane.org/handbook.
Derogatis 1974
- Derogatis LR, Lipman RS, Rickels K, Uhlenhuth EH, Covi L, Pichot P. The Hopkins Symptom Checklist (HSCL). Behavioral Science 1974;19:1-15. [DOI] [PubMed] [Google Scholar]
DerSimonian 1986
- DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clinical Trials 1986;7(3):177-88. [DOI] [PubMed] [Google Scholar]
Dias 2010
- Dias S, Welton NJ, Caldwell DM, Ades AE. Checking consistency in mixed treatment comparison meta-analysis. Statistics in Medicine 2010;29(7-8):932-44. [DOI] [PubMed] [Google Scholar]
Dionne 2008
- Dionne CE, Dunn KM, Croft PR, Nachemson AL, Buchbinder R, Walker BF, et al. A consensus approach toward the standardization of back pain definitions for use in prevalence studies. Spine (Phila Pa 1976) 2008;33(1):95-103. [DOI] [PubMed] [Google Scholar]
Distiller SR [Computer program]
- DistillerSR. Ottawa, ON: DistillerSR Inc (developed by Evidence Partners), 2022. Available from www.distillersr.com/products/distillersr-systematic-review-software.
Dunn 2006
- Dunn KM, Croft PR. The importance of symptom duration in determining prognosis. Pain 2006;121(1-2):126-32. [DOI] [PubMed] [Google Scholar]
Egger 1997
- Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315(7109):629-34. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
Endnote 2017 [Computer program]
- EndNote X8. Version X9. Philadelphia, PA: Clarivate, 2017. Available at www.endnote.com.
EuroQol 2019
- EuroQol Research Foundation. Q-5D-5L User Guide. Available from euroqol.org/publications/user-guides 2019.
Fairbank 1980
- Fairbank JC, Couper J, Davies JB, O’Brien JP. The Oswestry low back pain disability questionnaire. Physiotherapy 1980;66(8):271-3. [PubMed] [Google Scholar]
Ferreira 2013
- Ferreira ML, Herbert RD, Ferreira PH, Latimer J, Ostelo RW, Grotle M, et al. The smallest worthwhile effect of nonsteroidal anti-inflammatory drugs and physiotherapy for chronic low back pain: a benefit-harm trade-off study. Journal of Clinical Epidemiology 2013;66(12):1397-404. [DOI] [PubMed] [Google Scholar]
Furlan 2015
- Furlan AD, Malmivaara A, Chou R, Maher CG, Deyo RA, Schoene M, et al. 2015 Updated method guideline for systematic reviews in the Cochrane Back and Neck group. Spine 2015;40(21):1660-73. [DOI] [PubMed] [Google Scholar]
GBD 2017
- GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet 2018;392(10159):1789-858; Erratum in: Lancet. 2019 Jun 22;393(10190):e44. [DOI: 10.1016/S0140-6736(18)32279-7] [DOI] [PMC free article] [PubMed] [Google Scholar]
Ghogomu 2014
- Ghogomu EA, Maxwell LJ, Buchbinder R, Rader T, Pardo Pardo J, Johnston RV, et al. Updated method guidelines for Cochrane musculoskeletal group systematic reviews and metaanalyses. Journal of Rheumatology 2014 Feb;41(2):194-205. [DOI] [PubMed] [Google Scholar]
Grabovac 2019
- Grabovac I, Dorner TE. Association between low back pain and various everyday performances. Wiener klinische Wochenschrift 2019;131(21):541-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hamilton 1986
- Hamilton M. The Hamilton rating scale for depression. In: Assessment of Depression. Springer, 1986:143-52. [Google Scholar]
Harris 2019
- Harris PA, Taylor R, Minor BL, Elliott V, Fernandez M, O’Neal L, et al. The REDCap consortium: building an international community of software partners. Journal of Biomedical Informatics 2019;95:1-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hartvigsen 2018
- Hartvigsen J, Hancock MJ, Kongsted A, Louw Q, Ferreira ML, Genevay S, et al. What low back pain is and why we need to pay attention. Lancet 2018;391(10137):2356-67. [DOI] [PubMed] [Google Scholar]
Hayden 2009
- Hayden JA, Chou R, Hogg-Johnson S, Bombardier C. Systematic reviews of low back pain prognosis had variable methods and results: guidance for future prognosis reviews. Journal of Clinical Epidemiology 2009;62(8):781-96. [DOI] [PubMed] [Google Scholar]
Hayden 2020
- Hayden JA. Predatory publishing dilutes and distorts evidence in systematic reviews. Journal of Clinical Epidemiology 2020;121:117-9. [DOI] [PubMed] [Google Scholar]
Hayden 2021c
- Hayden JA, Ellis J, Ogilvie R, Boulos L, Stanojevic S. Meta-epidemiological study of publication integrity, and quality of conduct and reporting of randomized trials included in a systematic review of low back pain. Journal of Clinical Epidemiology 2021;134:65-78. [DOI] [PubMed] [Google Scholar]
Hayden 2022
- Hayden JA, BACK Evidence Collaboration - Collaborative Review Working Group. Collaborative systematic review may produce and share high-quality, comparative evidence more efficiently. Journal of Clinical Epidemiology 2022 [Epub 2022 Sep 28];152:288-94. [DOI: 10.1016/j.jclinepi.2022.09.013] [DOI] [PubMed] [Google Scholar]
Hays 2009
- Hays RD, Bjorner JB, Revicki DA, Spritzer KL, Cella D. Development of physical and mental health summary scores from the patient-reported outcomes measurement information system (PROMIS) global items. Quality of Life Research 2009;18(7):873-80. [DOI] [PMC free article] [PubMed] [Google Scholar]
Heitz 2009
- Heitz CA, Hilfiker R, Bachmann LM, Joronen H, Lorenz T, Uebelhart D, et al. Comparison of risk factors predicting return to work between patients with subacute and chronic non-specific low back pain: systematic review. European Spine Journal 2009;18(12):1829-35. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hellerstein 2008
- Hellerstein JM. Quantitative data cleaning for large databases. United Nations Economic Commission for Europe (UNECE) 2008;25:1-42.
Higgins 1996
- Higgins JP, Whitehead A. Borrowing strength from external trials in a meta-analysis. Statistics in Medicine 1996;15(24):2733-49. [DOI] [PubMed] [Google Scholar]
Higgins 2002
- Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Statistics in Medicine 2002;21(11):1539-58. [DOI] [PubMed] [Google Scholar]
Higgins 2011
- Higgins JP, Altman DG, Gøtzsche PC, Jüni P, Moher D, Oxman AD, et al, Cochrane Bias Methods Group, Cochrane Statistical Methods Group. The Cochrane Collaboration's tool for assessing risk of bias in randomised trials. BMJ 2011;343:1-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
Higgins 2022a
- Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.3 (updated February 2022). Cochrane, 2022. Available from www.training.cochrane.org/handbook.
Higgins 2022b
- Higgins JPT, Li T, Deeks JJ, editor(s). Chapter 6: Choosing effect measures and computing estimates of effect. In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.3 (updated February 2022). Cochrane, 2022. Available from www.training.cochrane.org/handbook.
Higgins 2022c
- Higgins JPT, Eldridge S, Li T, editor(s). Chapter 23: Including variants on randomized trials. In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.3 (updated February 2022). Cochrane, 2022. Available from www.training.cochrane.org/handbook.
Hozo 2005
- Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Medical Research Methodology 2005;5:1-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
Itz 2013
- Itz CJ, Geurts JW, Kleef M, Nelemans P. Clinical course of non-specific low back pain: a systematic review of prospective cohort studies set in primary care. European Journal of Pain 2013;17(1):5-15. [DOI] [PubMed] [Google Scholar]
Jonge 2013
- Jonge ED, Loo MV. An introduction to data cleaning with R. Statistics Netherlands, 2013. [Google Scholar]
Kopec 2000
- Kopec JA. Measuring functional outcomes in persons with back pain: a review of back-specific questionnaires. Spine 2000;25(24):3110-4. [DOI] [PubMed] [Google Scholar]
Kroenke 2001
- Kroenke K, Spitzer RL, Williams JB. The PHQ‐9: validity of a brief depression severity measure. Journal of General Internal Medicine 2001;16(9):606-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lim 2012
- Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H, et al. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 2012;380(9859):2224-60. [DOI] [PMC free article] [PubMed] [Google Scholar]
Martin 2008
- Martin BI, Deyo RA, Mirza SK, Turner JA, Comstock BA, Hollingworth W, et al. Expenditures and health status among adults with back and neck problems. JAMA 2008;299(6):656-64. [DOI] [PubMed] [Google Scholar]
McGowan 2016
- McGowan J, Sampson M, Salzwedel DM, Cogo E, Foerster V, Lefebvre C. PRESS peer review of electronic search strategies: 2015 guideline statement. Journal of Clinical Epidemiology 2016;75:40-6. [DOI] [PubMed] [Google Scholar]
Melzack 1975
- Melzack R. The McGill Pain Questionnaire: major properties and scoring methods. Pain 1975;1:277-99. [DOI] [PubMed] [Google Scholar]
Miyamoto 2019
- Miyamoto GC, Lin CC, Cabral CMN, Dongen JM, Tulder MW. Cost-effectiveness of exercise therapy in the treatment of non-specific neck pain and low back pain: a systematic review with meta-analysis. British Journal of Sports Medicine 2019;53(3):172-81. [DOI] [PubMed] [Google Scholar]
National Guideline Centre 2016
- National Guideline Centre (UK). Low Back Pain and Sciatica in Over 16s: Assessment and Management. London: National Institute for Health and Care Excellence (NICE); 2016 Nov. (NICE Guideline, No. 59). Available from www.ncbi.nlm.nih.gov/books/NBK401577/. [PMID: ] [PubMed]
NCCDPHP 1993
- National Center for Chronic Disease Prevention and Health Promotion. CDC HRQOL–14 "Healthy Days Measure". Available from www.cdc.gov/hrqol/hrqol14_measure.htm 1993.
Nikolakopoulou 2020
- Nikolakopoulou A, Higgins JP, Papakonstantinou T, Chaimani A, Del Giovane C, Egger M, et al. CINeMA: an approach for assessing confidence in the results of a network meta-analysis. PLoS Medicine 2020;17(4):e1003082. [DOI] [PMC free article] [PubMed] [Google Scholar]
Noel‐Storr 2021
- Noel-Storr A, Dooley G, Elliott J, Steele E, Shemilt I, Mavergames C, et al. An evaluation of Cochrane Crowd found that crowdsourcing produced accurate results in identifying randomized trials. Journal of Clinical Epidemiology 2021;133:130-9. [DOI] [PubMed] [Google Scholar]
Page 2021
- Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. [DOI] [PMC free article] [PubMed] [Google Scholar]
Powell 2011
- Powell KE, Paluch AE, Blair SN. Physical activity for health. What kind? How much? How intense? On top of what? Annual Review of Public Health 2011;32:349-65. [DOI] [PubMed] [Google Scholar]
Qaseem 2017
- Qaseem A, Wilt TJ, McLean RM, Forciea MA. Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American College of Physicians. Annals of Internal Medicine 2017;166(7):514-30. [DOI] [PubMed] [Google Scholar]
R Foundation for Statistical Computing [Computer program]
- R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2013. Available at www.R-project.org.
Radloff 1977
- Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Applied Psychological Measurement 1977;1(3):385-401. [Google Scholar]
Retraction Watch 2023
- The Center for Scientific Integrity. The Retraction Watch Database. Available from retractiondatabase.org/ (accessed 15 May 2023).
RevMan Web [Computer program]
- Review Manager Web (RevMan Web). Version 5.2.0. The Cochrane Collaboration. Available at revman.cochrane.org.
Riley 2011
- Riley RD, Higgins JP, Deeks JJ. Interpretation of random effects meta-analyses. BMJ 2011;342:d549. [DOI] [PubMed] [Google Scholar]
Roland 2000
- Roland M, Fairbank J. The Roland-Morris Disability Questionnaire and the Oswestry Disability Questionnaire. Spine (Phila Pa 1976) 2000;25(24):3115-24. [DOI] [PubMed] [Google Scholar]
Shafshak 2021
- Shafshak TS, Elnemr R. The Visual Analogue Scale versus Numerical Rating Scale in measuring pain severity and predicting disability in low back pain. Journal of Clinical Rheumatology 2021;27(7):282-5. [DOI] [PubMed] [Google Scholar]
Slade 2016
- Slade SC, Dionne CE, Underwood M, Buchbinder R. Consensus on Exercise Reporting Template (CERT): explanation and elaboration statement. British Journal of Sports Medicine 2016;50(23):1428-37. [DOI] [PubMed] [Google Scholar]
Song 2003
- Song F, Altman DG, Glenny AM, Deeks JJ. Validity of indirect comparison for estimating efficacy of competing interventions: empirical evidence from published meta-analyses. BMJ 2003;326(7387):472-72. [DOI] [PMC free article] [PubMed] [Google Scholar]
Thomas 2021
- Thomas J, McDonald S, Noel-Storr A, Shemilt I, Elliott J, Mavergames C, et al. Machine learning reduced workload with minimal risk of missing studies: development and evaluation of a randomized controlled trial classifier for Cochrane reviews. Journal of Clinical Epidemiology 2021;133:140-51. [DOI] [PMC free article] [PubMed] [Google Scholar]
Tukey 1977
- Tukey JW. Exploratory data analysis. Addison-Wesley, 1977. [Google Scholar]
Von Korff 2000
- Von Korff M, Jensen MP, Karoly P. Assessing global pain severity by self-report in clinical and health services research. Spine 2000;25(24):3140-51. [DOI] [PubMed] [Google Scholar]
Ware 1992
- Ware JE, Sherbourne CD. The MOS 36 item short-form health survey (SF-36). 1: conceptual framework and item selection. Medical Care 1992;30:473-80. [PubMed] [Google Scholar]
WHOQOL Group 1998
- WHOQOL Group. Development of the World Health Organization WHOQOL-BREF quality of life assessment. Psychological Medicine 1998;28(3):551-8. [DOI] [PubMed] [Google Scholar]
Wiklund 1990
- Wiklund I. The Nottingham Health Profile: a measure of health-related quality of life. Scandinavian Journal of Primary Health Care 1990;1:15-8. [PubMed] [Google Scholar]
Wood 2020
- Wood L, Ogilvie R, Hayden JA. Specifying the treatment targets of exercise interventions: do we? British Journal of Sports Medicine 2020;54(20):1235-6. [DOI] [PubMed] [Google Scholar]
Wood 2021
- Wood L, Bishop A, Lewis M, Smeets RJ, Bronfort G, Hayden JA, et al. Treatment targets of exercise for persistent non-specific low back pain: a consensus study. Physiotherapy 2021;112:78-86. [DOI] [PubMed] [Google Scholar]
Wu 2020
- Wu A, March L, Zheng X, Huang J, Wang X, Zhao J, et al. Global low back pain prevalence and years lived with disability from 1990 to 2017: estimates from the Global Burden of Disease Study 2017. Annals of Translational Medicine 2020;8(6):1-14. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wun 2021
- Wun A, Kollias P, Jeong H, Rizzo R, Cashin AG, Bagg MK, et al. Why is exercise prescribed for people with chronic low back pain? A review of the mechanisms of benefit proposed by clinical trialists. Musculoskeletal Science and Practice 2021;51:102307. [DOI] [PubMed] [Google Scholar]
Yepes‐Nuñez 2019
- Yepes-Nuñez JJ, Li SA, Guyatt G, Jack SM, Brozek JL, Beyene J, et al. Development of the summary of findings table for network meta-analysis. Journal of Clinical Epidemiology 2019;115:1-13. [DOI] [PubMed] [Google Scholar]
Zigmond 1983
- Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatrica Scandinavica 1983;67(6):361-70. [DOI] [PubMed] [Google Scholar]
Zung 1986
- Zung William WK. Zung self-rating depression scale and depression status inventory. In: Assessment of Depression. Springer, 1986:221-31. [Google Scholar]
References to other published versions of this review
Hayden 2005a
- Hayden JA, Van Tulder MW, Malmivaara A, Koes BW. Exercise therapy for treatment of non-specific low back pain. Cochrane Database of Systematic Reviews 2005, Issue 3. Art. No: CD000335. [DOI: 10.1002/14651858.CD000335.pub2] [DOI] [PMC free article] [PubMed] [Google Scholar]
Hayden 2005b
- Hayden JA, Van Tulder MW, Malmivaara AV, Koes BW. Metaanalysis: exercise therapy for nonspecific low back pain. Annals of Internal Medicine 2005;142(9):765-75. [DOI] [PubMed] [Google Scholar]
Hayden 2012
- Hayden JA, Cartwright J, Van Tulder MW, Malmivaara A. Exercise therapy for chronic low back pain. Cochrane Database of Systematic Reviews 2012, Issue 4. Art. No: CD009790. [DOI: 10.1002/14651858.CD009790] [DOI] [PMC free article] [PubMed] [Google Scholar]
Hayden 2021a
- Hayden JA, Ellis J, Ogilvie R, Malmivaara A, Tulder MW. Exercise therapy for chronic low back pain. Cochrane Database of Systematic Reviews 2021, Issue 9. Art. No: CD009790. [DOI: 10.1002/14651858.CD009790.pub2] [DOI] [PMC free article] [PubMed] [Google Scholar]
Hayden 2021b
- Hayden JA, Ellis J, Ogilvie R, Stewart SA, Bagg MK, Stanojevic S, et al. Some types of exercise are more effective than others in people with chronic low back pain: a network meta-analysis. Journal of Physiotherapy 2021;67(4):252-62. [DOI] [PubMed] [Google Scholar]