Assessing the impact of manual therapy in the management of nonspecific low back pain: a scoping review of the outcomes used in clinical trials

Abstract

Background

Low Back Pain (LBP) is the leading cause of disability worldwide, 90% of which is nonspecific. Manual therapy is one of the recommended treatment modalities. However, reported outcomes may be variable. This review aims to identify their scope in the context of the development of a Core Outcome Set (COS), which is defined as « an agreed standardised set of outcomes that should be measured and reported, as a minimum, in all clinical trials in specific areas of health or health care ».

Methods

A scoping review with risk of bias assessment of randomised controlled trials (RCTs) of manual therapy for nonspecific LBP was conducted using MEDLINE, CENTRAL, PEDro, WebOfScience and ClinicalTrials.gov, from 2010 up to August 2024. Manual therapy was considered the use, alone or in combination, of manipulations (high velocity, low amplitude), mobilisations (low-grade velocity, small-to-large amplitude) or soft tissue relaxation (especially massage, trigger points, muscle contractions).

Results

Out of 3929 articles, 147 RCTs and 74 protocols were included. Two main outcomes emerged: pain intensity (assessed by numerical rating scale or visual analogue scale) and disability (mostly assessed by Rolland-Morris Disability Questionnaire or Oswestry Disability Index). Range of motion is the most frequent clinical outcome assessed. Psychological factors such as fear-avoidance beliefs, kinesiophobia and catastrophising, and healthcare consumption, particularly medication, are also frequent. Most of the outcomes were patient-reported outcomes.

Conclusion

Consistent with a previous COS on nonspecific low back pain, manual therapy appears to address the same outcomes. Clinical trials in manual therapy should focus on using the existing COS by measuring pain intensity using a numerical rating scale, disability using the ODI 2.1a or the 24-item RMDQ, health-related quality of life using the SF-12 or the 10-item PROMIS. Additionally, due to the gap between clinical research and pain experience, trials should consider conducting subgroup analyses to identify effects on outcomes related to gender or age, paying particular attention to health inequalities by carrying out analyses based on socioeconomic status, as these factors are well known to significantly impact pain experience and access to care.

Review protocol

PROSPERO registration CRD42024576475, COMET Database registration 3229.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12906-026-05288-4.

Background

Low back pain (LBP) is a highly prevalent disease, and the main cause of disability worldwide, with an estimated point prevalence of 10.4% (CI95%: 10.0–10.9) and a lifetime prevalence ranging from 65 to 80% [1, 2]. It is estimated that up to 95% of LBP cases are nonspecific, meaning that no anatomopathological cause can be found to explain the pain [3]. This is supported by several studies exploring the prevalence of imaging findings in healthy patients. These studies reveal that pathological signs are more prevalent than assumed in general population, regardless of the clinical context [4, 5]. However, numerous factors influence its development, including smoking, obesity, lack of physical activity [6], with an increased risk of chronicity after acute episodes, particularly in contexts involving emotional distress (e.g. anxiety or depression), inappropriate behaviours and attitudes (e.g. fear-avoidance beliefs or kinesiophobia) and occupational issues [7, 8]. For this reason, many national and international guidelines recommend an early intervention, including non-pharmacological interventions, such as rehabilitation and manual therapy interventions [8–12]. Notably, low back pain is the main reason for seeking manual therapy [13–15]. However, evidence synthesis can be impaired by factors including the heterogeneity of outcomes (criteria) and outcome measures (tools used to measure the criteria) used in clinical research [12, 16–18], or a weak methodological quality, particularly when the intervention is compared to a sham [19–21], directly because of a poorly described intervention [22] or due to difficulties in blinding [23, 24]. Recommendations have been made to improve the trustworthiness and quality of clinical trials, in osteopathy [25] and of non-pharmacological interventions in general [26]. A recent step forward by the Non-Pharmacological Intervention Society (NPIS) highlights the need to define relevant outcomes, either in effectiveness clinical trials (CI6: « Define the main health objective and the primary outcome », CC14: « Specify secondary objectives » and CO22: « Choose relevant outcomes measured by validated and sensitive tools ») and prototype study (PO9: « Define in advance the main health outcome which the NPI prototype is supposed to improve »). Such initiatives imply the development of a Core Outcome Set (COS), which is defined as « an agreed standardised set of outcomes that should be measured and reported, as a minimum, in all clinical trials in specific area of health or health care » [27]. The development of a COS is a key part of such initiatives to develop evidence-based consensus on relevant outcomes and outcome measures. These initiatives emerged in the early 90’s [28] and structured themselves around a standardised methodology to identify the need for a COS, establishing it and achieving similar processes for outcome measures [27, 29, 30]. This review is the first step toward establishing the need of a COS for the management of nonspecific LBP with manual therapies.

Methods

A scoping review was performed as recommended for this type of question [27, 31]. The implementation and choices made at each stage are detailed in the following paragraphs, particularly with regard to the availability of team members.

Selection criteria

We searched a wide range of databases to include a large variety of randomised clinical trials of manual therapy for the treatment of nonspecific low back pain, including sacroiliac joints. No restrictions were placed on the type of control or comparator. Interventions should be:

• Manipulations (high-velocity, low amplitude) applied on joints on the spine or elsewhere,

• Mobilisations (low-grade velocity, small to large amplitude) applied on joints on the spine or elsewhere,

• Soft tissue relaxation (massage, trigger points or muscle contractions against resistance).

The following databases were used: CENTRAL, Cochrane Library, PEDro, Web of Science Core Collection and ClinicalTrials.gov, from 2010 to august 2024, the extraction date. To provide a comprehensive review of the existing literature, including articles that may have been incorrectly referenced, we chose to use both MeSH and non-MeSH terms. We retrieved the full text of each article. To include contemporary trials, we also searched ClinicalTrials.gov for ongoing trials. Articles including children (aged 18 years or younger) and those published in languages other than English or French were excluded. Study protocols were included if they described a clinical trial that was eligible for inclusion but not yet published. On PubMed, the following research equation was used: “Clinical Trial” [Publication Type] AND ((“Musculoskeletal Manipulations“[Mesh] OR chiropract*[Title/Abstract] OR osteopath*[Title/Abstract] OR “manual therapy“[Title/Abstract] OR manipulat*[Title/Abstract]) AND (“Back Pain“[Mesh] OR “Low Back Pain“[Mesh] OR “Neck Pain“[Mesh])). Full search strategy is available as a supplementary file.

Data extraction

The following characteristics were extracted for each study: first author, year of publication, participating countries, sample size, mean age of participants, study duration, type of intervention, primary outcome, other outcomes and outcome measures, including the rational for their use. Outcome measures were defined as any measure (including events) that was reported for all study arms. Where reported, the following data were extracted: outcome domain (e.g. physical function), outcome measure (e.g. PROMIS short form v.2.0 physical function 4a), and the time points at which measurements were taken during the trial. Finally, risk of bias was assessed using the recommended tool for RCTs, Cochrane Risk of Bias 2.0. Selection was performed using Rayyan [32]. An automation tool (NotebookLM) was used for a quick first extraction of article information, which was then reviewed and corrected by the authors.

Analysis

Four authors (MB, JP, BD and YS) were responsible for selecting trials. In case of disagreement, a fifth author (MS) would resolve the issue. Author MB categorised all domains into three categories: clinical (medical events that occur as a result of disease or treatment, e.g. adverse events or medication consumption), patient-reported (outcomes that are directly reported by patients regarding how they feel or function, e.g. pain intensity or satisfaction), and surrogate (markers that are used as a proxy, validated or not, for clinical outcomes, e.g. tissue stiffness) [33–35]. The number of trials reporting each outcome was calculated. Secondary analyses included outcomes used depending on LBP duration. Outcome measures will be presented for the main outcomes used. The results were then compared with existing core outcome sets to estimate the usefulness of a core outcome set dedicated to nonspecific low back pain treated with manual therapy and to compare the types of outcomes retained. Analyses and data visualisations were made using R 4.4.3 [36].

Protocol and deviations

This scoping review has been registered on PROSPERO (CRD42024576475) and the COMET Database (3229). However, several protocol deviations occurred, including:

• Focusing on nonspecific low back pain instead of back pain in general, due to the researchers’ time constraints. This project is still ongoing,

• Focusing on articles from 2010 instead of not applying a time limit, as this is the most suitable time frame used for developing COS,

• Focus on RCTs, as these are the main design for effectiveness trials and help to standardise data extraction and reduce the number of risk of bias tools used,

• The method for aggregating criteria was not extracted in order to reduce the workload, given its lack of relevance at this stage of the COS development process,

• Due to the small number of criteria compared to other COS, detailed analyses of the most frequent criteria by domain have been omitted as they did not provide additional information,

• Further analyses, especially based on country or date of publication, were discarded due to the consistency of the results.

Particularly, the review selected articles with the original inclusion criteria until full-text screening. This review follows the PRISMA Extension for Scoping Reviews checklist [37].

Results

The flow diagram is available in Fig. 1. The screening strategy extracted 221 articles, of which 147 published RCTs and 74 protocol of unpublished studies, ongoing or not. The Fig. 2 shows the distribution of the timeframe of declaration or publication, depending on the type of report. Risk of bias assessment is available in a supplementary file. Of the included trials, 69 (47%) showed a low risk of bias, 40 (27%) presented some concerns and 38 (26%) were at high risk.

Fig. 1.

PRISMA flow diagram

Fig. 2.

Number of protocols (n = 74) and published articles (n = 147) depending on the year of declaration or publication respectively

Of the 217 outcomes used, only 43 were used more than twice. These are shown in Fig. 3. Most of these outcomes (n = 22, 51%) were patient-reported, 14 (33%) were clinical and the remaining 7 (16%) were surrogate. The two most frequently reported outcomes were pain intensity (n = 187), which was assessed using numerical rating scale or visual analogue scale, and disability (n = 184), which was assessed using Oswestry Disability Index (ODI), Rolland-Morris Disability Questionnaire (RMDQ), or occasionally other questionnaires. Range of motion was assessed with multiple tools, including goniometers or inclinometers, or using clinical tests such as finger-floor distance. General health included several questionnaires, such as SF-36 and SF-12, EQ-5D and PROMIS-29. Pressure pain threshold was only assessed using an algometer. Fear-avoidance beliefs and kinesiophobia were evaluated using the Fear-Avoidance Beliefs Questionnaire and Tampa Scale of Kinesiophobia, respectively. Taken altogether, care consumption (drug use, use of care) was also frequent (n = 31). Other outcome measures and frequency of use are described in Table 1, full results are available in a supplementary file.

Fig. 3.

Overall outcomes that appear more than twice (n = 43) among the included articles (n = 221), log10 scale

Table 1.

Outcomes and related outcome measures (n > 2). Empty cells relate to unspecified, rarely used or unapplicable measures. ^†Frequency of outcome measure may not match frequency of outcome if measures were unspecified or used less than 3 times

Outcome	Frequency of outcome	Outcome measure	Frequency of outcome measure†
Pain intensity	187	NRS	102
		VAS	85
Disability	184	ODI	105
		RMDQ	72
		QBPDI	4
Range of motion	63	Schober	18
		Finger-floor distance	16
		Inclinometer	14
		Goniometer	12
General health	58	SF-36	27
		SF-12	11
		EQ-5D	10
		PROMIS29	8
Pressure pain threshold	38	Algometer	38
Fear-avoidance beliefs	30	FABQ	30
Kinesiophobia	26	Tampa	26
Perceived change	23	GROC	16
Depression	21	BDI	8
		HADS	5
		PHQ	3
Drug use	21
Adverse events	19
Satisfaction	19
Catastrophizing	14	PCS	14
Endurance	13	Biering-Sorensen	5
		McQuade test	4
Pain experience	13	MPQ	11
Anxiety	11	HADS	5
Global improvement	11
Postural sway	11	Stabilometer	5
		Motion analysis	3
		Baropodometer	3
Use of care	10
Self-efficacy	9	PSEQ	6
Strength	9	Dynamometer	3
Electromyographic activity	8	EMG	8
Risk of chronicisation	8	STarT Back	6
Flexibility	7	Sit and reach test	6
Stiffness	7	Myotonometer	5
Fall risk	6	Timed up and go test	4
Kinematics measures	6	Motion analysis	6
Pain frequency	6
Thickness	6	Ultrasound	6
Balance	5
Bothersomeness	5
Specific function	5	PSFS	5
Temporal summation	5	Von Frey filaments	5
Acceptable symptom state	4	PASS	4
Back performance	4	Back performance scale	4
Central sensitization	4	CSI	4
Number of lost activity days	4
Sleep quality	4	PSQIQ	4
Treatment expectations	4
Multifidus activation	3	EMG	3
Physical function	3
Recovery	3
Sensory discrimination	3	Esthesiometer	3

Figure 4 provides subgroup analyses depending on the duration of LBP. Regardless of pain duration, pain intensity and disability remain the most used outcomes. General health (health-related quality of life) and fear-avoidance beliefs are also frequently assessed regardless of duration. Care consumption (drug use, use of care) and adverse events are consistently assessed in approximately 10–15% of clinical trials, while range of motion and pressure pain threshold are more frequently evaluated as the duration of LBP increases. Although not often assessed, a total of 8 times, the risk of chronicisation is evaluated in cases of chronic LBP, 6 times.

Fig. 4.

Outcomes used in trials that appear more than twice depending on the duration of LBP among the trials. Acute trials (n = 62) reported 20 outcomes. Subacute trials (n = 64) reported 24 outcomes. Chronic trials (n = 193) reported 41 outcomes. Trials may include various durations

The time points used to assess the outcomes are shown in Fig. 5. Most of endpoints were short- to mid-term, ranging from 3 weeks to 3 months. 28 outcomes were only assessed immediately after one intervention. 34 assessments took place after 6 months, including 28 endpoints. The longest follow-up took place 3 years after the study began.

Fig. 5.

Time points used as follow-up (left) or as end of follow-up (right)

Discussion

This review is the first to provide a comprehensive view of outcomes used in clinical trials of manual therapy targeting nonspecific low back pain. It reveals that, although only 43 outcomes are used more than twice, a wide range are employed. More than 80% of trials reported at least pain intensity or disability, which is consistent with a close COS for nonspecific low back pain (regardless of the therapy), that identified pain intensity, physical functioning, health-related quality of life and number of deaths as core outcomes [38]. Although the outcomes are named differently between articles, this COS defines physical functioning as the « impact on patients’ ability to carry out daily physical activities required to meet basic needs (…) », which is the construct assessed by the RMDQ and ODI [39]. Health-related quality of life is also the same construct as this reviews’ general health domain. However, the number of deaths is never assessed in manual therapy clinical trials. Conversely, range of motion is the third most frequently used outcome in manual therapy clinical trials, yet it is not part of the published COS. Similarly, pain pressure threshold, fear-avoidance beliefs and kinesiophobia are not discussed in the COS despite being common outcomes measured in manual therapy trials. This may be related to the lack of use of this COS or a lack of interest for these outcomes when using manual therapy as an intervention. Consensus has already been reached on the tools to be used for these outcomes, with the NRS for pain intensity, ODI 2.1a or 24-item RMDQ for disability and SF-12 or 10-item PROMIS for general health [40]. These were the tools most frequently identified in this review, except for pain intensity for which almost 50% of studies used visual analogue scale. The SF-36 was also the most frequently used questionnaire for assessing general health in this review, despite not being the recommended one [40]. It should be noted that the widespread use of the main measurement tools may hinder the development of other tools, particularly for complex outcomes, since ensuring comparability with previous studies is important. This is noticeable in the case of disability and general health, which can be assessed using various questionnaires. While a COS proposes a minimum set of outcomes, research teams should not refrain from exploring other aspects of the outcomes, including core ones, if relevant. As illustrated by Fig. 5, most of the trials focus on short to mid-term timepoints, with various follow-up intervals. Although easier to obtain, focusing on the rapid effects of treatment can obscure the long-term effects, whether beneficial or undesirable. In addition, assessing short-term effects is also becoming less relevant given the large number of clinical trials focusing on the treatment of chronic pain. Although there is no consensus on how frequently follow-ups should be conducted during the first year for non-radicular spinal pain, recent recommendations offer guidance on the frequency and duration of follow-ups in chronic pain treatment. These recommendations suggest follow-ups every one or two months during the first year and then every six months thereafter [41].

Strength and limitations

The main strength of this review is that it provides a systematic analysis of contemporary literature on manual therapy, focusing on the manual components of care rather than professional categories, as is often the case. The collective involvement of four authors in the selection of articles aimed to minimize bias as much as possible. However, the variety of the interventions considered in the review may have made synthesising of the results reported by each author more difficult, as descriptions may vary from one person to another. Supervision of this process by one author (MB) aimed to reduce this disparity. Nevertheless, it cannot be ruled out that certain outcomes or outcome measures were reported incorrectly in some cases, despite the manual review of the results being intended to prevent this. Although all stages were discussed during meetings, several were carried out by a single author due to availability issues, particularly with regard to the analyses, the production of figures, and additional documents. The aim of this process was to comply with the recommendations for conducting a scoping review as much as possible [31, 42].

The main limitation is the diversity of interventions, which makes it difficult to conduct a detailed analysis of the outcomes used for each type of intervention. This is because trials often combine several treatment modalities (manipulation, mobilisation, soft tissue relaxation), or compare them in different groups. However, it is assumed that such an analysis would not reveal any significant disparities. As this review did not focus on quantitative synthesis of the results, it is not possible to draw conclusions about the effect of the interventions. This was not the objective of this review. It should be noted that of the included articles, 28 protocols were submitted before 2020, with no published article since, some of which were submitted almost 15 years ago. It is possible that some articles have been published in languages or in journals that could not be identified. This nevertheless raises questions about the transparency of the results and the proper use of the resources dedicated to conducting these clinical trials.

Further considerations

Beyond the need of a COS, several studies have shown that poor psychological factors, including fear-avoidance beliefs, kinesiophobia and catastrophizing, are highly associated, especially in chronic low back pain, with greater pain intensity and disability, and poorer recovery status of acute low back pain [43–45]. This is supported by recent guidelines [8, 9, 46]. Therefore, manual therapy has been shown to have a small to moderate effect on these factors [9]. As manual therapy is a non-pharmacological treatment modality, it could help reduce care consumption, particularly with regard to drug prescription, and more specifically with regard to opioids and other drugs with a risk of dependence. Several large-scale retrospective cohorts appear to show a significant effect on the prescription and consumption [47–49], consistently with a recent systematic review [50].

Furthermore, some of the included studies stratified their randomisation either by age (n = 7), gender (n = 14) or pain characteristics (n = 11), or a combination of these factors. This is justified by the fact that these factors are correlated with different pain experiences that may confound the results [1]. However, almost none of the included studies reported subgroup analyses based on these factors, to identify differences in the response to care. Some articles provided secondary analyses in another article [51] while others dealt with potential factors using multivariate analyses [52]. Therefore it has been shown that women are more likely to experience pain, and, on the other hand, are less likely to receive appropriate care [53–55]. Beyond being mistrusted with pain being minimised, women experiencing pain tends to be more often referred to psychological care [55]. Thus, it would be appropriate to encourage subgroup analyses based on gender, in order to identify any gender-specific effects or mechanisms of treatment. On the other hand, some research suggest that certain outcomes, such as pressure pain threshold [56] and pain perception in general [57, 58] are affected by ageing. Without subgroup or sensitivity analyses based on age or age categories, a pooled analysis of results may mask highly variable realities depending on age. Conducting such analyses could enhance the interpretation of the results and their subsequent impact on current practice, if any. Although taking these factors into account is necessary for randomisation, it is insufficient to identify potential specific effects. It is therefore well known that health inequalities affect disadvantaged socio-economic groups more than other, particularly with regard to access to care, consideration of health status, and the quality of care pathways [59]. This socio-economic status is sometimes reported. However, the most disadvantaged socioeconomic groups are less included in health research, with only few studies dedicating subgroup analyses to them [60, 61]. These remarks on gender, age, and socioeconomic status are not mutually exclusive, with elderly and poor women being a particularly vulnerable population group, among others. Of course, carrying out subgroup analyses requires a certain sample size to ensure the results are interpretable and to avoid statistical artefacts. The issue of sample size is a recurring problem in manual therapy trials, as it is in trials of other therapies [12, 62, 63]. These analyses also present other statistical challenges [64].

Conclusion

Manual therapy seems to rely on relevant outcomes that are shared with other therapies for the management of nonspecific low back pain. Manual therapy trials should focus on using the core outcomes already identified, along with the recommended outcome measures: pain intensity measured using numerical rating scale, disability or physical function measured using Oswestry Disability Index 2.1a or 24-item Rolland-Morris Disability Index, general health or health-related quality of life measured with SF-12 or 10-item PROMIS and number of deaths. Beyond the core outcomes, manual therapy research uses a wide range of additional outcomes that may be relevant in certain cases, such as range of motion, pressure pain threshold and care consumption. Clinical research should encourage subgroup analyses to identify any gender- and/or age-specific effects and better take into account the socioeconomic status of vulnerable populations, particularly those who meet several of these characteristics.

Abbreviations

COS

Core outcome set

LBP

Low back pain

NPIS

Non-pharmacological intervention society

ODI

Oswestry disability index

RMDQ

Rolland-Morris disability questionnaire

Authors’ contributions

Mathis BRIER developed the protocol, conducted the article selection and screening, data extraction, data synthesis and analyses and wrote the manuscript. Jules PHALIP provided feedback on the protocol and conducted screening. Benjamin DELORME and Yves SCHEWENDENMANN conducted screening. Maxime SALMON managed the selection conflicts. Dominique J. BICOUT provided feedback on the protocol and helped writing the manuscript. Christine ROLLAND provided feedback on the protocol and helped writing the manuscript. Julien NIZARD provided feedback on the protocol and helped writing the manuscript. Each author has approved the manuscript and accepted responsibility for their contributions to this work.

Funding

This project is funded by IdHEO Nantes, School of Osteopathy, which covers the salary of Mathis BRIER. IdHEO Nantes did not contribute to the methodological consideration and had no influence on any stage of the work.

Data availability

Data is provided within the manuscript or supplementary information files. R code for generating the figures is available on request.Supplementary material 1 provides the full search strategy for each database that was considered.Supplementary material 2 contains the complete results of extracted informations used in the analyses. Supplementary material 3 provides the PRISMA Checklist for Scoping Reviews for the current article. Supplementary material 4 provides the full risk of bias assessment of the included studies.

Declarations

Ethical approval and consent to participate

Not applicable – scoping review.

Consent for publication

Not applicable – no data from any individual.

Competing interests

This project is funded by IdHEO Nantes, School of Osteopathy, which covers the salary of Mathis BRIER. IdHEO Nantes did not contribute to the methodological consideration and had no influence on any stage of the work.Mathis BRIER, Jules PHALIP, Benjamin DELORME, Yves SCHWENDENMANN, Maxime SALMON are osteopaths. Maxime SALMON is employed by IdHEO Nantes. Julien NIZARD is responsible for a DIU in manual medicine - osteopathy.