Manual Therapy Effect in Placebo-Controlled Trials: A Systematic Review and Meta-Analysis

Purpose: Background: Evaluate whether the design of placebo control groups could produce different interpretations of the efficacy of manual therapy techniques. Methods: Nine databases were searched (EMBASE, CINAHL, PsycINFO, MEDLINE, PubMed, SCOPUS, WEB of SCIENCE, COCHRANE, and PEDro). Randomized placebo-controlled clinical trials that used manual therapy as a sham treatment on subjects suffering from pain were included. Data were summarized qualitatively, and meta-analyses were conducted with R. Results: 53 articles were included in the qualitative analysis and 48 were included in the quantitative analyses. Manipulation techniques did not show higher effectiveness when compared with all types of sham groups that were analyzed (SMD 0.28; 95%CI [−0.24; 0.80]) (SMD 0.28; 95%CI [−0.08; 0.64]) (SMD 0.42; 95%CI [0.16; 0.67]) (SMD 0.82; 95%CI [−0.57; 2.21]), raising doubts on their therapeutic effect. Factors such as expectations of treatment were not consistently evaluated, and analysis could help clarify the effect of different sham groups. As for soft tissue techniques, the results are stronger in favor of these techniques when compared to sham control groups (SMD 0.40; 95%CI [0.19, 0.61]). Regarding mobilization techniques and neural gliding techniques, not enough studies were found for conclusions to be made. Conclusions: The literature presents a lack of a unified placebo control group design for each technique and an absence of assessment of expectations. These two issues might account for the unclear results obtained in the analysis.


Introduction
Pain is not merely an accompanying symptom to a pathology, but rather a separate condition on its own [1], that has such a profound effect on the individuals that commonly ends up leading to depression [2]. Millions of Europeans [3,4] suffer from acute and chronic pain, resulting in a significant physical, emotional, and financial burden [5]. Therefore, there is an urgent need for optimal management.
Among the therapies used to manage musculoskeletal pain, manual therapy is one of the most common approaches. Manual therapy can be described as the application of a manual force applied accurately and specifically to the body, to improve pain-related symptoms and mobility in areas that are restricted or injured such as joints, connective tissues, or skeletal muscles [6]. It comprises a group of therapeutic techniques including, amongst others, soft tissue techniques (i.e., ischemic compression, pressure inhibition), neural mobilization, joint mobilization, and manipulation which can modulate pain, especially in a short time [7][8][9]. On the other hand, neural mobilization or neurodynamic techniques are used for analgesic purposes and are intended to improve adaptability, reduce mechanosensitivity and activate analgesic mechanisms through mechanical stimulation of the nerves through elongation and sliding [10,11]. Alternatively, pursuing the same aim, passive specific joint mobilizations can be carried out at low velocity and high amplitude of movement or high velocity and low amplitude of movement (manipulation) of articulation [12]. Passive-specific mobilizations at low velocity and high amplitude of movement consist of rhythmic, passive, and smooth movements on a joint, where strength and amplitude are controlled according to the tolerance level of the patient [13]. The manipulation technique, on the other hand, consists of making a fast and controlled movement at the end of the range of a joint, producing cavitation [14].
Although manual therapies are commonly effective in the management of musculoskeletal conditions; several authors [15][16][17] believe that their positive effects result from the placebo effect, and not necessarily from their supposed therapeutical effect. The placebo effect is defined as a genuine positive psychological and/or physiological effect, occurring in humans and other animals, attributable single-handedly to the knowledge of receiving a substance or undergoing a procedure, but not due to the inherent power of that substance or procedure [18]. Placebo effects can be generated by previous experience, perception, information acquired about the treatment, and by active integration of all of these with the sensory information patients learn during a treatment [19,20]. Thus, in the handling of pain, the placebo effect is especially important. Nevertheless, it is a combination of such physical discomfort associated with an emotional response to it, that depends on previous experience, cultural background, emotions, and other neurobiological variations, exactly in the same way as the placebo effect does [21].
When using sham treatments as control groups to determine the effectiveness of a certain active therapy, such as manual therapies, it is important to design them well. They must be non-active and harmless actions that by themselves should not produce any physical improvement/deterioration of the patient's condition, residing here the importance of including a proper sham for each and every one of the evaluated manual therapies. Moreover, the blinding of the patients must be ensured to render a proper and adequate placebo intervention and reliable results [22,23].
Placebo-controlled Randomized Clinical Trials (RCTs) are studies that use a sham control group to simulate the treatment to be evaluated. As already stated, the purpose of this group is to provide a reliable comparator, so that the effect of a substance (i.e., placebo) or procedure (i.e., sham) can be exclusively attributed to itself [24,25]. In addition to highquality sham control, other methodological factors such as the blinding of both placebo and treatment groups (regardless of the subjects beliefs [26]), randomization allocation concealment, and adequate control of biases are of importance when determining the accuracy of treatment effects. From a methodological point of view, when an RCT does not guarantee that the sham control group considers all non-specific effects of the treatment, fails to ensure that the placebo is inert, or fails to confirm that its control subjects were efficiently blinded, it makes it difficult to make confident interpretations of the results of the trial [26]. Therefore, it is important to determine whether the sham groups used in the literature have been designed to provide reliable comparisons. However, we could not identify any previous review regarding this subject.
Thus, the objective of the study was to evaluate whether the design of placebo control groups could produce different interpretations of the efficacy of manual therapy techniques.

Materials and Methods
The systematic review and meta-analysis were carried out following the Cochrane Handbook for Systematic Review of Intervention [27] and the guide "Preferred Reporting Items for Systematic Reviews, PRISMA" [28,29]. The protocol was registered at PROSPERO with ID: CRD42020157468.

Literature Search
The following databases were searched until 11 December 2019, from inception: EM-BASE, CINAHL, PsycINFO, MEDLINE, PubMed, SCOPUS, WEB of SCIENCE, COCHRANE, and PEDro. The search strategy is based on the combination of medical terms (Mesh) and keywords, regarding the following scheme: Population: >18 years old subjects; Intervention: manual therapy; Comparative group: placebo OR sham; Outcome: pain scale. Subsequently, the search results were managed in Mendeley, proceeding with duplicate removal.

Study Selection
• Types of Studies. Parallel RTCs carried out in humans were included, however, crossover designs were excluded to avoid the carry-over effect. In addition, letters, reports, and abstracts from congresses were also excluded. • Types of population. All studies which were conducted with subjects suffering from pain were included, regardless of the characteristics of the participants. Nevertheless, when the pain was artificially induced in the subjects (i.e., experimental pain models), studies were excluded. • Types of interventions. It was compulsory for the studies to be included, to have used as a unique technique one of the following manual therapy techniques: manipulation, mobilization, soft tissue techniques, and neural mobilization in at least one intervention group. Intervention groups that were complemented with other techniques (i.e., electrotherapy, exercise) were not included, to allow the analysis of the individual effect of each technique. In the same way, therapies that involved active subject participation (i.e., exercise) were excluded. Finally, alternative medicine therapies such as reflexology or holistic treatments were excluded. • Types of comparison. To be included in the systematic review, the studies had to include at least one sham group, termed by the authors as: a placebo group, or sham group. If the subjects did not receive any intervention in the placebo or sham group, it was not considered a valid sham comparison group. • Types of outcomes. Pain intensity had to be assessed with rating scales such as the Visual Analog Scale (VAS), Numerical Rating Scale (NRS), Numerical Pain Rating Scale (NPRS), or similar. • Time points. For the quantitative analysis, the pre-treatment assessment and the first available post-treatment assessment were chosen.

Selection of Studies
Firstly, duplicates were removed and transferred to a Microsoft Excel spreadsheet; afterward, articles were screened through title and abstract for eligibility, and finally, a full-text evaluation of the studies was performed. Every article identified in the different databases was reviewed independently by two authors (C.R.-R. and A.A.-R.). The authors had to reach a consensus on whether a trial should be included. If the authors did not reach a consensus, a third author (M.M.-A.), who participated in the process, made the final decision.

Data Extraction
Required data was extracted by two independent reviewers (C.R.-R. and A.A.-R.) using an excel sheet. From the included studies, data such as author(s), year of publication, participant characteristics (sample size, gender, age, and pathology), group description, outcome measurements, and results, were collected. After data compilation, both authors had to reach a consensus on each item of the extracted data, and in case of disagreement between the authors, a third author (M.M.-A.) made the final decision.

Blinding Quality Assessment
Since the aim of the study was to evaluate whether the design of placebo control groups could produce different interpretations of the efficacy of manual therapy techniques, the blinding of the subjects was assessed. Assessment of blinding quality was performed through the Risk of Bias (RoB) of the Cochrane Handbook for Systematic Reviews of Interventions to assess bias [30], independently by two reviewers (C.R.-R. and A.A.-R.). Any discrepancies in quality ratings were solved by discussion. If consensus could not be reached, a third member of the review team (M.M.-A.) acted as an arbitrator.
The analysis of blinding was complemented, as other authors have described [31], with the evaluation of other domains of the RoB tool such as random sequence generation, allocation concealment, participant blinding, and detection bias. The overall RoB for the above key domains was named as: "Adequate" if a trial was rated low risk (all key domains as low risk or at most with one unclear risk domain); or "Not adequate" if a study was considered an unclear risk (unclear risk in more than one domain) or high risk (high risk in at least one domain). Furthermore, research personnel blinding and therapist blinding proved to be impossible to achieve due to the techniques procedure, therefore these domains were not taken into account.
All the studies which did not perform a simulation of the intervention groups (i.e., fake massage as a control of manipulation interventions) were cataloged as high risk in "participant blinding". Hence, studies that applied detuned devices or simulations of other techniques were assumed to not be valid. We considered that these placebo control groups could interfere with the expectations of the participants, patients undergoing a manual therapy technique do not have the same experiences as patients undergoing an electrotherapy-based treatment. Placebo is defined as a genuine positive psychological and/or physiological effect, occurring when undergoing a procedure, but it is not due to the inherent powers of that procedure [18], thus, the placebo groups have to guarantee: the replication of the protocol, with the objective that the subjects have the same expectations and undergo the same subjective experience; harmlessness, the sham group should not produce a greater effect than the technique, at most, the same effect.
To assess inter-rater reliability, the kappa coefficient (κ) was calculated to assess reliability prior to any consensus and the inter-rater reliability was estimated by using the kappa coefficient (κ) considering that κ > 0.7 indicates a high level of agreement among reviewers; κ of 0.5-0.7 indicates a moderate level of agreement; and a κ < 0.5 indicates a low level of agreement [32]. Disagreements on the quality assessment of RoB were resolved by consensus with a third independent reviewer.

Data Analysis and Synthesis
Data was analyzed qualitatively and compiled in evidence tables. The quantitative analyses were conducted with the metafor package in R [33,34]. The results were analyzed using the Standardized Mean Difference (SMD) with its 95% Confidence Interval (CI). Since different scales were used to measure pain, the SMD Hedge's g was used. All forest plots reported results on pain intensity. The effect size was evaluated following Cohen's magnitude criteria for rehabilitation treatment effects: d = 0.14-0.31 "small" effect size; d = 0.31-0.61 "medium" effect size; and d > 0.61 "large" effect size [35]. Effects were interpreted as statistically significant when p < 0.05. The inverse variance statistical analysis method was used. Combined results analysis was carried out using the random effect model, as opposed to the fixed effect model, incorporating the variance in each study and between studies [27].

Assessment of Heterogeneity
Heterogeneity among trials was assessed using heterogeneity statistics (e.g., Cochrane's Q test and I 2 statistic). It was considered that Q p-values < 0.10 indicate a significant heterogeneity [35]. In addition, I 2 describes the percentage of variability due to heterogeneity, and not to random error or chance. I 2 values between 0% and 30% were considered to have null to low heterogeneity; between 30% and 50%, medium heterogeneity; between 50% and 75%, moderate heterogeneity; and over 75%, high heterogeneity [36].

Subgroup Analysis
After the overall analysis, subgroup analyses were performed according to blinding adequacy and type of control sham group. When looking at the adequacy of blinding, two groups were compared: studies with adequate blinding versus those with inadequate blinding. On the other hand, when studying the effect of types of sham groups, studies were grouped into the following categories: detuned devices, "therapeutic" (i.e., soft massage), and different types of simulation of the techniques (i.e., simulation with movement).

Assessment of the Level of Evidence
To assess the overall quality of evidence in the outcomes of the meta-analysis, the Grading of Recommendations Assessment, Development and Evaluation (GRADE) [37] approach was implemented. The quality of evidence was categorized as follows: high: the true effect lies close to the estimate of the effect; moderate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different; low: the true effect may be substantially different from the estimate of the effect; very low: the true effect is likely to be substantially different from the estimate of the effect.

Selection of Trials
The literature search produced 8753 results, of which 5163 were duplicated. After the first screening, a total of 3091 studies were excluded. Following the whole text reading, 445 articles were excluded and 53 remained. Finally, these 53 articles were included in the qualitative analysis; of which only 48 were included in the quantitative analysis, as they did present the necessary measures to be analyzed ( Figure 1). Inter-rater reliability for assessing the risk of bias was very high (κ = 0.94). The principal characteristics of the trials are available in Table 1.

VAS Preintervention
Postintervention Cyclic and rhythmic mobilizations were applied to the talus in the anteroposterior direction.
The therapist simulated the treatment, giving a light force on the joint, but did not actually produce movement.

Risk of Bias' Blinding Assessment
A summary of the results of the RoB analysis is displayed in Table 2. For the adequacy of blinding, the overall results were assessed. The trials published by Atkinson et al. [52], Pikula et al. [40], and Lehtola et al. [64] were classified as high risk in the key domain participant blinding since these studies did not simulate the intervention group procedure. In these designs, the subjects could be blinded on whether they underwent a manipulation treatment but were not blinded in the procedure of the manipulative technique. Several trials (21%) presented high risk in the detection bias domain [39,40,46,53,54], being cataloged as "Not adequate" for not meeting the required criteria. Finally, 19 trials (79%) accomplished the requirements to be considered as adequately blinded for the study group [38,41,[43][44][45][47][48][49]51,52,[55][56][57][58][59]62,63,66,67]. ? ?
x Yes Hondras et al. [
x No

Quantitative Analysis
Overall meta-analysis showed statistical significance in favor of manipulation interventions, nevertheless moderate heterogeneity was presented (SMD 0. 42  When the results of RoB were introduced in the subgroup analysis, statistical significance was found (Q = 4.38; p = 0.036), furthermore within the group cataloged as "Not adequately" blinded, the results did not find statistical significance between the intervention group and placebo control group. On the other hand, the adequately blinded group did present differences in favor of manipulation ( Figure 2). In the last subgroup analysis, analyzing different types of sham groups, the results showed no statistical significance differences between groups (Q = 1.88; p = 0.599). Nevertheless, it is worth mentioning, that in the forest plot analysis the only group that presented statistical differences was the simulated manipulation one, without the movement of the joints. But no differences were found in either the group that simulated applying movement, the device, or the interventions which seemed to be therapeutic (Figure 3).

Placebo
As in the manipulation techniques studies, the authors followed mainly the same two strategies. In the simulation procedure, the authors who tried to simulate the technique mainly chose the gently positioning of the hands on the massage area in the absence of movement or pressure (28.6%) [73,79,82,83]; the other simulation (7.1%) consisted of faking the same treatment, applying superficial massage on the structures trying to minimize, to the biggest extent, contact with the skin [78]. Detuned devices were used as an alternative sham group, the authors applied detuned ultrasound (21.5%) [72,76,77], detuned microwave diathermy (7.1%) [74], algometer (7.1%) [75], and detuned transcutaneous electric nervous stimulation (7.1%) [81]. Three studies neither simulated procedures nor used detuned devices: Antolinos-Campillo et al. [71] performed flexion and extension of the hip for four min as a sham group against suboccipital muscle inhibition in subjects with cervical whiplash; Celenay et al. [80] performed unspecific strokes and effleurage in the lower back area, and Espí-López et al. [61] performed the artery test as placebo control for tension-type headache.

Risk of Bias Blinding Assessment
A summary of the results of RoB is exposed in Table 2. The minimum requirements for adequately blinding the subjects were achieved by five trials (35.7%) [73,[78][79][80]82]. On the other hand, nine trials were cataloged as "Not adequate" for the final analysis since six studies were cataloged as high risk in the participant blinding domain (42.8%) [71,72,74,76,77,81], given the subjects were not blinded for the soft tissue techniques procedure. High risk in allocation concealment was presented by Tanaka et al. [83], and finally, Espí-López et al. [61], was categorized as an unclear risk in more than one key domain.

Quantitative Analysis
When the meta-analysis was performed including all the studies, medium effect size, statistical significance differences and low heterogeneity were found (SMD 0. 40  No statistical difference was found between the adequate blinding studies and the not adequate blinding trials (Q = 0.11; p = 0.74). However, the trials cataloged as not adequately blinded showed statistical differences between techniques and placebo control groups, and the adequate blinding studies did not show statistical differences in favor of techniques ( Figure 4). No statistical difference was found between types of sham control groups (Q = 1.19; p = 0.55). Moreover, the studies that performed placebos with possible therapeutic effects did not show differences between soft tissue techniques and sham interventions. When laying hands were applied or detuned devices were used, statistical differences in pain were found between "active" and sham groups. (Figure 5).

Participants
A total of 544 subjects were included in manual therapy studies in which, at least, one intervention group applied isolated mobilization techniques. 327 subjects of the total subjects were females, and the average age of all the participants was 41.29 years old. Regarding the prevalence of pathologies in the subjects: two studies (25%) evaluated the neck pain [43,84], two (25%) experimented in patients suffering of low back pain [85,86], and ankle pain (12.5%) [87], cervico-craniofacial pain (12.5%) [88], thoracic pain (12.5%) [89] and shoulder pain (12.5%) [90] were studied each in one trial.

Intervention Groups
The most common intervention (50%) consisted of one treatment session [43,84,89,90], however, La Touche et al. [88] performed six sessions in three weeks and Kogure et al. [85] carried out six sessions in six months. Regarding the mobilization emplacements, the authors mainly (75%) conducted the interventions on the vertebral spine [43,[84][85][86]88,89]. Furthermore, the doses, in terms of application time and the number of repetitions, were diverse.

Placebo
The authors for the design of sham mobilizations mostly reproduced the intervention protocol but used light force in the application of the technique with the objective of not producing a therapeutic effect on the subjects (50%) [85,87,89,90] or mimicking the grip and the procedure of mobilization without performing any force (25%) [86,88]. Both sham procedures reproduced the intervention protocol. On the contrary, Snodgrass et al. [84] did not reproduce the mobilization protocol and chose a detuned ultrasound as the sham group, and Valera-Calero et al. [43] used as the sham group, the sham manipulation group and not the sham mobilization intervention.

Risk of Bias' Blinding Assessment
A summary of the RoB results is available in Table 2. Only two studies (25%) succeeded in terms of adequately blinding [88,89]. Snodgrass et al. [84] was cataloged as "Not adequate" due to the impossibility of participant blinding since the study did not reproduce the mobilization procedure, the same as Valera-Calero et al. [43]. Also, three studies (37.5%) were cataloged as high risk in the detection bias domain [85][86][87]. Eventually, Surenkok et al. [90] also presented unclear risks in three of the four key domains.

Quantitative Analysis
In the overall analysis, the meta-analysis did not show statistically significant difference between mobilization techniques and sham control groups (SMD 0.

Participants
From the four identified trials, a total of 341 subjects were recruited, 294 of which were females. The average age of all participants was 40.1 years old. Regarding the pathologies: two trials (50%) proved the effectiveness of neural gliding in carpal tunnel syndrome [91,92], one trial (25%) studied subjects who presented neck pain [93], and the last one (25%), included subjects who suffered from tension-type headache [78].

Intervention Groups
The most common neurodynamic technique applied was neural gliding on the median nerve (75%) [91][92][93], however, the authors applied different dosifications according to the time of application and number of sets. Fernández-Carnero et al. [93] conducted only one session, while the other studies carried out more than one intervention session, with the Wolny et al. [92] trial being the longest, with 20 sessions in 10 weeks. Only Ferragut-Garcías et al. [78] performed neural gliding techniques in a different localization, specifically, on the cranio-cervical area.

Placebo
All trials (75%), except for Ferragut-Garcías et al. [78], mimicked the intervention group in the same way, trying not to apply tension to the median nerve but reproducing to the maximum extent the neural mobilization technique. Ferragut-Garcías et al. [78] instead reproduced the sham neurodynamic technique by simulating the soft tissue technique intervention also presented in the trial.

Risk of Bias' Blinding Assessment
A summary of the RoB results is available in Table 2. The trial of Ferragut-Garcías et al. [78] was the only one that was cataloged as inadequately blinded since they did not secure the blinding participant because they did not reproduce the neurodynamic technique procedure. The rest of the trials (75%) were cataloged as successful overall for adequately blinding [91][92][93].

Quantitative Analysis
The meta-analysis of the studies included did not show statistical differences and high heterogeneity was found (SMD 0. 99  . Since only four studies were included, it was not possible to perform the subgroup analysis (Figure 7).

Quality of Evidence
After the assessment of the quality of evidence through GRADE, the level of recommendation for manipulation techniques and soft tissue techniques was low for the management of pain. In the case of mobilization techniques and neurodynamic techniques it was very low (Table 3).

Discussion
The results revealed that it is not clear whether manual therapy techniques are superior to sham control groups. The main reason may reside in the use of multiple designs of sham groups found in the literature. When different sham groups were divided into subgroups, less heterogeneity was found, and some conclusions could be made.
Throughout the last decades, the placebo effect has been the subject of numerous studies by the scientific community. However, in the field of manual therapy, the placebo effect has been undervalued on numerous occasions, probably as a result of the difficulty of blinding the participants, the lack of reliable interventions, and the inability of blinding the personnel (responsible for applying the interventions). In 2017, Bialosky et al. [94] questioned the effectiveness of the techniques of manual therapy and how many of those possible outcomes were due to the placebo effect. The evaluation of blinding success (for the subject or evaluator) has been a requirement of guidelines to design quality sham groups [95][96][97]. The blinding index has proved to be an interesting tool to properly assess blinding in RCTs. Furthermore, incorporating this tool in manual therapy studies [98,99] could provide factual information on the perception of the subjects who underwent a research intervention. Regarding the blinding of the personnel, it is convenient to note that although this factor could influence the outcomes of the studies [100], in this analysis it was assumed that this requirement in manual therapy trials would be impossible to achieve.
Nonetheless, it would be important to point out that sham groups are not widespread and standardized in the literature. Manipulation is the technique most researchers use to seek plausible sham groups; authors such as Vernon et al. [44,70], Chaibi et al. [101], and Michener et al. [69,102] have designed protocols for the development of valid sham groups. In this regard, the results are along the same lines as other reviews that evaluated sham groups [19,22]. Vernon et al. [24] and Puhl et al. [26] assessed the quality of sham groups in trials of cervical manipulation, and lumbar and pelvic manipulation, respectively, both systematic reviews concluded the lack of quality sham control groups. Metanalyses were also carried out assessing the different types of placebo control groups for manipulation techniques in chronic back pain [103] and chronic low back pain [104], nevertheless, contradictory results were found. In addition, Hancock et al. [25] questioned 25 experts on which sham procedure could be more feasible to control manipulation RCTs, but amongst the participants, there was an extremely low level of agreement.
When comparing the studies according to the adequacy of blinding evaluated by the RoB, active manipulation techniques showed significantly bigger pain reduction when compared to sham groups in trials with adequate blinding. Nevertheless, in the subgroup analysis involving several types of sham, the only studies which simulated the technique without applying movement-active treatment showed significantly better pain reduction than the sham group. The studies which chose a detuned device or simulated the highspeed movement did not show statistical differences between active and sham groups. This contradiction could be explained by the expectations generated by the different control groups; in other words, it is possible that the positive results of the manipulation in pain management may be caused by the positive expectations of the subjects.
Contrary to the last analysis, and also beholding RoB analysis, in the soft tissue techniques, the studies that were adequately blinded did not show statistically significant differences between active and sham groups, while studies that were inadequately blinded showed a statistical difference in favor of soft tissue techniques. However, when the analysis was carried out in the function of the types of sham control groups, the control groups which had potential therapeutic effects did not show statistical differences, but the detuned device group and the simulation group did. Therefore, soft tissue techniques presented positive effects on pain management. Recent reviews share similar results in terms of the effectiveness of myofascial release [105] and ischemic compression [106] techniques in the management of myofascial pain syndrome.
As for the mobilization techniques and the neurodynamic techniques, the limited number of studies constrained our analysis. No significant differences were found between groups for any of the analyses performed, nor were we able to perform the analysis in the function of the types of placebos. Further studies comparing neurodynamic techniques and mobilization are needed. Recent investigations have found that neural mobilization is useful to manage neck pain [107] and back pain [107,108], however higher quality RCT could be clarifying.
The aim of placebo control groups is to generate the same expectations in the subjects as the expectations generated by the "active" intervention groups [18].
Also, it is worthwhile to highlight that all these results support the exclusion from this work of studies that used detuned devices as placebo groups for not implementing a reliable sham control group for manual therapy techniques. Nevertheless, when analyzing the results from different sham procedures, detuned devices obtained similar results as the other sham procedures. For future research, it could be interesting to evaluate the expectations of the different sham control groups and how that affects treatment effect estimates.

Study Limitations
The study presents several limitations. Firstly, the lack of studies found that compared manual therapy with sham control groups. It may be due to the difficulty and the low consensus when authors have to develop quality placebos. Furthermore, another limitation is the length of the interventions, since the number and the frequency of the sessions were not taken into consideration, causing dropouts was not discussed, as in previous studies [109]. Nor were the characteristics of the participants such as sex, age, or pain etiology compared. Lastly, only studies written in English were included. The effect of blinding was evaluated following similar guidelines as previous studies [31]; however, blinding could be analyzed in different ways and thus future research should challenge this evaluation and perhaps perform sensitivity analyses for different types of blinding when using sham groups.

Conclusions
In summary, the literature presents a lack of a unified placebo control group design for each technique and an absence of expectations assessed. These two issues might account for the unclear results obtained in the analysis. The manipulation techniques were demonstrated to be more effective in pain reduction than placebo control groups in the overall analysis. However, manipulation techniques did not show superior effectiveness when compared with all types of placebos, raising doubts about their therapeutic effect which could be resolved in future studies with the evaluation of participants' expectations in the different sham groups, therefore the efficacy could not be concluded in this study. In the case of soft tissue techniques, the results are stronger in favor of these techniques when compared to placebo control groups, being the soft tissue techniques more effective than placebo control groups for the management of pain. Even so, the authors suggest the same recommendations, the evaluation of participants' expectations. Finally, regarding the mobilization techniques and neural gliding techniques, not enough studies were found in order to make conclusions.