The effect of spinal manipulative therapy on experimentally induced pain: a systematic literature review

Pain is defined as an unpleasant sensory and emotional experience associated with actual or potential tissue damage[1]. It originates in specific receptors, named nociceptors, which are classified according to the type of damage to which they respond; thus, mechanosensitive, thermosensitive, chemosensitive and polymodal nociceptors.

From the peripheral nociceptors, noxious stimuli are transmitted to the dorsal horn of the spinal cord[2]. There, afferent fibers synapse in the superficial laminae of the dorsal gray matter of the spinal cord[3]. Cells in the superficial laminae serve as an integration centre and relay system for many sensations. Most cells of the grey matter involved in nociception send axons across the midline of the spinal cord to ascend to the thalamus. From there, they project upwards, eventually to the cortex of the brain. However, impulses are not only transmitted to the cerebral cortex. There are also other pathways and mechanisms that may participate in analgesic influences at the spinal and supraspinal levels.

When a noxious stimulus occurs, there may be a “first pain”, conducted by Aδ fibers, and a “second pain”, due to temporal sensory summation (TSS) and conducted by nociceptive C-fibers. “First pain” is described as sharp and "pricking". The propagation of this stimulus is relatively quick and it is felt in a well-defined part of the body surface[2]. “Second pain”, which is transmitted more slowly, is often described as dull and aching, and it is poorly localized. This pain tends to last beyond the termination of an acute noxious stimulus. Sources, pathways, perception of and treatment of the two types of pain are very different[4].

The infliction of pain is not always experienced in a linear manner according to the strength and nature of the stimulus. Pain sensations can be very different from one individual to another, and also intra-individual variations can occur, so that identical types of damage do not necessarily result in an identical amount and type of pain. One reason for this is that pain can be modulated, both to increase and decrease.

One modulating system, central sensitization, tends to increase pain sensation[5], particularly in people who have more long-lasting pain, making them more sensitive to “new” pain impulses than they would have been under normal circumstances.

Another endogenous modulating system is afferent or segmental inhibition, meaning that one external stimulus can block an ongoing pain sensation by having higher priority in reaching the brain[4].

Descending antinociceptive systems provide yet another modulating mechanism. These originate largely in the mesencephalon and have synaptic connections with neurons in the medulla and the spinal cord. This means that nociceptive information may be blocked or attenuated before it reaches higher centers[6]. This system is also tightly connected to a descending pain facilitating pathway that has the same general sites of origin (mesencephalon and medulla) but with the opposite effect.

Finally, there are also other intrinsic mechanisms for physiological modulation of pain, such as subjective assessment and motivational-affective modulation[7], which act by raising pain thresholds via endogenous opioids and other substances. These mechanisms, at times, preferentially alter sensory and/or affective aspects of pain perception, and the associated modulation of pain-evoked neural activity occurs in limbic and/or sensory brain regions, suggesting multiple endogenous pain-modulating systems[8]. Thus, pain can be increased or decreased by mere expectations or, even, abolished by feelings of, for example, fear.

Although back pain is common and frequently distressing, and many therapies have arisen to treat it, musculoskeletal pain remains difficult to diagnose and treat. Spinal manipulative therapy (SMT) is one common treatment for musculoskeletal pain. One class of SMT involves a high-velocity, low-amplitude (HVLA) manipulation frequently used by chiropractors. HVLA treatments are mechanical events. They cause slight momentary deformations of the spine and surrounding soft tissues, and often elicit a cracking sound thought to be brought about by cavitation of spinal facet joints[9‐11]. It is common to differentiate manipulation from mobilization. In the latter case, the joint is not taken beyond its passive limit. Rather mobilization can be described as a passive and perhaps repetitive stretch. Manipulation, on the other hand, carries the vertebrae beyond the normal physiological range of movement without exceeding the boundaries of anatomic integrity[11]. However, the distinction between manipulation and mobilization is probably not that clear, and it has been shown that cavitation is not necessary for SMT to exert a clinical effect[12‐16]. The term SMT can therefore be used to describe various types of manual therapy (MT).

Clinical experience indicates that both HVLA and mobilization, and also other types of manual therapy (MT), can have an immediate effect on pain. The literature also suggests that SMT has a direct neurological pain-reducing effect, by evoking one or possibly several of the physiologic pain-modulating mechanisms described briefly above. Indeed, there could be a combination of mechanisms or a number of these acting on different causes of pain. In this article we shall concentrate on the possible direct effect of SMT on pain. There are three possible levels of this hypothesized direct effect of SMT on pain, i.e. local, regional or central.

Also a comprehensive model of mechanisms of manual therapy in the treatment of musculoskeletal pain has been suggested[24] consisting of a cascade of neurophysiological responses from the peripheral and central nervous systems which are then responsible for the clinical outcomes. In other words, the literature offers many possible mechanisms and combinations of mechanisms to explain the pain-reducing effect of SMT. In fact, there is a lot of information available in the literature on this topic. However, the literature is difficult to grasp and conflicting because it consists of a mixture of discussions, hypotheses, and studies employing different designs, methods and outcome variables. Therefore, there is a need for systematic and critical literature reviews in order to establish the evidence-base for various theories relating to the direct or indirect reduction of pain following SMT. A first step might be to establish the weight of evidence in relation to whether pain is indeed dampened by the application of SMT to the spinal structures.

Therefore, in this systematic critical literature review we shall examine the effect of spinal manipulation on experimentally induced pain in healthy study subjects concentrating on the possible direct effects of SMT on pain at three levels, i.e. local, regional or systemic. Because different types of pain may travel through different pathways, these were studied separately.

The specific research questions were:

Each article selected was checked for each quality item. By consensus, we decided to give one point for every criterion, except for the 4 items that we considered as most important: a- unbiased/blind/naïve study subjects (item 1), b- random allocation (item 3), c- blinded assessment (item 9), and d- losses and exclusions (item 11). Each of these was assigned two points.

A summary quality score was calculated but no cutoff-point was defined for acceptable or unacceptable level of quality. This allowed us (and the readers) to use the quality scores and the information on each individual quality item as a guide for whether articles would be considered more or less credible. In other words, the quality assessment was meant to be flexible.

A post hoc comparison was made between results of the two quality scores (the total score and the score for the most important criteria). Using thresholds set at 12/18 and 4/8 points, respectively, the distributions of studies with positive and non-positive outcomes were compared for the two scales and in relation to whether the studies were of higher or lower quality.

The data tables were used, in a systematic fashion, to obtain answers to our research questions. This was done by highlighting relevant information to facilitate a visual representation (green = positive outcome, red = non-positive outcome) to make interpretation easier. Finally, studies were sorted in descending order starting with those with the highest total quality score out of 18. Results were thereafter interpreted and reported in a narrative fashion. If high and low quality studies generally obtained similar outcomes, we assumed that the poorer quality studies provided supporting evidence for the better studies. If, however, it was mainly the poorer quality studies which obtained positive findings, particularly if the assessors were not blinded or the study subjects could have been biased, then we would be more cautious in our interpretation of the results.

Because we decided not to use a scoring system to establish level of quality, we have not defined discrete levels of evidence.

Upon scrutinizing the full texts, 22 articles were found to fulfill the inclusion criteria. They described 43 experiments: 27 with pain produced by pressure, 9 by temperature, 3 by capsaicin, 2 examined spontaneous pain, 1 used a stretch test to produce pain and 1 used electrically induced pain. All were controlled trials published in English. Detailed information is presented in Table1 and briefly summarized below.

Three research groups were responsible for 13 publications. They did not seem to have repeatedly used the same study samples but appeared to have reported on different study populations for the various studies. In most cases (n = 18), experiments were carried out totally or partially in universities. The number of subjects ranged from 10 to 122. Six articles used patients as study subjects; most of the others included non-clinical populations, often students. There were no studies with animals fulfilling our inclusion criteria.

Most of experiments (n = 20) used external control groups. Twelve of them used three groups: a treatment, a placebo and a control group in which no action was taken - noted in tables as “nothing” (studies[21, 23, 28‐35]). Studies[18, 36] used bicycle exercise as their control activity. Seven studies ([19, 20, 37‐41]) used a sham procedure as their control treatment, whereas study[42] compared the treatment group to a control group without a sham procedure. Two studies used only internal control groups;[43] and[44] compared three different types of mobilization.

There were no disagreements between reviewers on the scoring of the manuscripts based on the checklists. Table2 summarizes the quality items for each article. The quality and risk of bias assessment of studies reviewed revealed them to be relatively homogeneous, with the summary scores ranging between 8 and 16 out of 18 points, with mean and median summary scores of 12.2 and 13, respectively.

Of the various quality items that we scrutinized, all studies fulfilled the criteria on the validity of the outcome variable, all had measured the pain before and after the intervention, all had provided some estimate of the results in tables, graphs or text, and all had tested their results for statistical significance. The ANOVA was used for factorial designs in all studies in order to test how variables interacted or combined.

Other checklist items were not always fulfilled. Of the four quality criteria that we considered particularly important, namely unbiased/blind/naïve study subjects (item 1), random allocation (item 3), blinded assessment (item 9), and losses and exclusions (item 11), the first two were commonly satisfied (in 15 and 17 papers, respectively), whereas using unbiased/blind/naïve study subjects and accounting for losses and exclusions were ignored in many studies (present in 7 and 5, respectively). Five of the reviewed articles ([18, 23, 36, 43, 44]) should be considered with care since their assessments of outcomes were not blinded.

Seven articles did not mention if the allocation of study subjects to each group was randomized. As it was not our intention to penalize unfairly articles that used a randomized allocation without mentioning it, we identified the numbers of individuals in each group in these reports (see Table3). Articles[19, 32, 33] did not provide the number of subjects in each experimental group. The others (except[30]) showed a more even distribution than one could expect with a proper randomization procedure. We therefore assumed that the fact that authors did not discuss the randomization procedure, in most cases, reflected the fact that it had not occurred.

In our review, we did not set a threshold for acceptable/unacceptable quality but displayed the articles in descending order from the highest to the lowest score. Nevertheless, in Table2a, we compared the spread of articles when using the total score (18 point scale, see Table2) and the scale based on four high-priority items (8 point scale, see Table2b). With levels of “acceptance” provisionally set at 12/18 and 4/8, respectively, all articles, with one exception, obtained the same classification of acceptable or not, regardless of the system used. We interpreted this concordance as a confirmation of the robustness of the quality and risk of bias assessment.

In addition to the quality issues identified in this review, the authors themselves have in some instances discussed the limitations of their own studies (see Table4).

A meta-analysis was not attempted because the heterogeneity of studies on pain provoking methods, units of measurements, areas of the body where experiments were performed, and local, regional of systemic assessments of the effects of SMT.

Herein, each research question is dealt with one at a time, and for each, we refer to a table summarizing the results in order of quality of the study. If one study contains several experiments, they have all been reported individually in the appropriate section of the review. Table5 presents studies grouped by their results, showing in the upper panels those that presented a hypoalgesic action and, in the lower panels, those that did not. Papers are also separated by the research question, i.e. local, regional or systemic effect. Table6 summarizes for each article the effect of SMT on pain, the site where SMT was performed and the location of the pain. This table also shows values reported, whether the effect was local, regional or systemic, whether the effects were ipsilateral or contralateral, whether an effect occurred above or below the site of manipulation, and the type of pain induced.

In the majority of experiments (28/43), SMT reduced pain. Outcomes were not dependent on whether quality scores were high or low. These studies demonstrate a clear hypoalgesic effect of SMT (see Table5a). For specific results, see below.

At the time of conducting this review, only one earlier systematic review was available on this subject. It examined 11 articles but was unable to draw conclusions regarding treatment effects[45]. For the present review, however, it was possible to identify 22 relevant articles, which made it possible to draw several conclusions in relation to the possible pain-reducing effect of SMT. Thus our results indicate clearly that such an effect is achievable.

However, differences in effect exist according to whether the outcome is tested locally, regionally or systemically. An effect was clearly shown locally and regionally, whereas an effect is less clear in more distant parts of the body. Also the outcome differs according to the method of pain induction; pain induced by pressure, electricity, stretching of painful tissue, dermal irritation, and spontaneous pain all respond to SMT, whereas temperature-induced pain does not always respond.

One limitation of this review is that there is no generally accepted and validated quality check list for the type of experimental studies which we examined and so we had to select our own quality criteria, based on some basic methodological concepts important to our research questions. Such a check list can be modified, which could affect the overall quality assessment. According to our quality scoring system, studies scored between 8 and 16 points out of a possible 18. However, in order to discriminate better between studies, a more detailed scale could have resulted in different scores, which could perhaps have separated studies into more obviously “good” and “bad”. However, the relevance, if any, of these quality items in relation to outcome, is not known and, in fact, the quality scores were not clearly associated with outcomes.

The strengths of this review are that the search for relevant articles was free of language bias, that the constituent articles were reviewed independently by two reviewers, that the results can be considered in relation to the quality of studies, that there was no arbitrary threshold for acceptable quality[46], that the check list tables are sufficiently detailed to allow readers to perform their own analysis of the information provided, and that there was a relatively large number of studies, making it possible to examine several research questions.

This review identified some common methodological problems, such as lack of blinded assessment that could weaken the evidence in these experimental studies. We considered one of the most important points to be that the assessor was blinded, to avoid expectation bias. Some studies ignored this issue. This aspect affected our interpretation of the systemic effect of SMT, as all five articles (10 trials) dealing with the systemic effects of SMT lacked a blinded assessor.

It is also important that the results be truthfully presented, and not exaggerated in some way, meaning that any subjects or data excluded from analysis should be accounted for. This was often ignored.

Another challenge was that of the control group. Ideally, the SMT should be matched against a suitable sham treatment and control procedure, and this was done in 12 studies. This is difficult with physical procedures, but several of the other 10 studies tried to make the best of the situation by selecting naive study subjects, and in some cases different types of treatment were compared (such as treatment in two areas of the spine), in which case no sham treatment would be necessary. However, whether a proper sham treatment was used or not, there were no obvious differences in the results, possibly indicating that the effect is very obvious and not affected to a large extent by expectations.

Potential confounders of these effects would be anxiety in general and fear of pain, known moderators of treatment outcomes in clinical practice. However, according to the three articles in which anxiety and fear of pain were studied, there were no statistically significant correlations between pain-related cognition, pain thresholds and the hypoalgesic response to SMT, indicating that the psychological aspect, considered to be so important in the perception of pain, perhaps does not come into play during experimental studies of this type.

Some of our results were corroborated by a newly published high quality review on this very topic[47]. This other review included 15 of our articles but, as they did not put a restriction on duration of symptoms in symptomatic people, they also incorporated 5 articles not included in our review. They defined effects as occurring “locally” or in “remote” areas, and pooled the results in all studies for the effects of PPT. They found the effects on PPT to be small in both areas but nevertheless statistically significant. When the data were separated for local and remote effect significance was only noted for the remote effect. They did not study any other pain–inducing method nor did they take into consideration the quality of studies, although they did perform a quality assessment.

Pain provoked by pressure, electricity, stretching of painful tissue, dermal irritation, or spontaneous pain all respond to SMT.

Experimental pain provoked by pressure was the most common method used to assess the effect of manipulation (27 out of 43 experiments). Unfortunately, the reports did not use the same units for reporting increased tolerance to pressure, making it difficult to compare outcomes, but SMT was shown to increase PPT values between 4.8% and 44.6%. That the algometer, used for this purpose, is a highly reliable method to measure pain[48] suggests that these results are robust. However, they cannot necessary be compared between studies because of different experimental situations.

Nevertheless, it is not clear whether these changes are clinically significant. One author[44] referred to Moss[49], who stated that a 15% improvement in pressure pain tolerance is clinically important, but this statement rests on previous studies which dealt with osteoarthritic pain[50] and pain in an emergency service[51], not experimental pain. Therefore, the cut-off point at which the reduction of artificially induced pain has reached a clinically significant level is not known.

In addition, it cannot be concluded that the effect of SMT, as seen in these studies, would be as large in "naturally" painful areas. Therefore, it is not certain that the degree of pain reduction detected in experimental studies can be extrapolated to other situations.

Apart from pressure and temperature, other sources of pain were: capsaicin to irritate the skin followed by skin stroking to evoke allodynia, hyperalgesia induced by mechanical means, and spontaneous pain. In these instances also, SMT was generally able to reduce pain.

Pain provoked by temperature does not always respond to SMT.

Nine trials from five papers examined the effects of SMT on temperature induced pain. In 3 of the 6 tests examining second pain, an effect was found but none of the 3 trials evaluating first pain could demonstrate such an effect. This indicates that SMT may have an effect on C-fiber mediated “second pain” but not on the more acute “first pain” mediated by A-delta fibers. This finding may help to resolve the mechanisms by which SMT reduces pain.

How broadly does the effect extend?

Almost all authors discussed the possibility of a systemic effect of SMT but this was tested in only 5 studies (9 experiments). The results were clearly positive, although two of the authors concluded that local effects were stronger than more distant ones. However, none of these studies used a blinded assessor, and so the reported results must be treated with caution.

Nine experiments out of the ten that studied the regional effect were positive (see Table5). However, in the reviewed articles, it was difficult to differentiate between strictly local vs. more regional effects. The definition of regional is uncertain if following the dermatomes, as they are known to differ from the text-book mappings[52] and the origins and distributions of cutaneous nerves differ from person to person. Thus, a negative outcome in one person may simply be due to that person having an unusual pattern of nerve distribution. Despite this, most studies showed a positive effect with relatively few study subjects. This could indicate either that negative results were removed from the analyses or that the effect is consistent.

Imprecise dermatomal mapping makes it difficult not only to separate clearly local from regional effects but also regional from systemic effects. Interestingly, the application of spinal manipulation at the atlanto-occipital region was found to have an effect on the masseter muscle, despite that muscle being supplied by the trigeminal nerve, which does not exit in the upper cervical spine. We therefore described this experiment under “regional” rather than “local” but it could perhaps equally well have been described it as “systemic”.

Concerning the action of SMT on pain produced above or below the manipulated segment, only a few studies dealt with this and no consistent findings emerged. It would be necessary to test this issue specifically, in order to work out whether manipulation-induced impulses travel up or down the spinal cord or whether both occur.

Five studies investigated whether the hypoalgesic effect was mainly on the side of the manipulation or if it also appeared on the opposite side. These studies consistently demonstrated a bilateral effect.

Some clinical concepts

Although this review was based on experimentally induced pain, four findings emerged that could have a bearing on clinical practice, or at least on the concepts on which clinical practice is based. Manipulation of a “restricted motion segment”, sometimes referred to as a “manipulable lesion”, was not required for the “treatment” to have an effect. Additionally, the side of the manipulation was irrelevant. However, the results from this review can only be considered from a clinical point of view, if SMT has a similar effect on pain in a clinical context, which was not investigated in this study.

Finally, although a hypoalgesic effect was shown, it is not known how long this effect lasts; long enough for a person with pain to be able to regain a normal movement pattern or only long enough to give an impression of improvement? Other effects of the SMT were not studied in this review; effects such as improved biomechanics or reduced inflammation of disturbed tissues[53].

There is a need to establish a coordinated global research strategy on the subject of SMT and pain relief. It seems unnecessary to conduct further research on simply whether SMT has an effect on pain or not. Rather, future work should focus on more precise questions such as why SMT does not affect first pain, and which mechanisms are involved in relieving pain locally, regionally and systemically. The magnitude and duration of effects also need to be defined. Consideration needs to be given to the most appropriate research designs for addressing different questions.

There could also be more consistency in outcome measures, and closer attention should be paid to important design elements such as blinded assessment, random allocation, appropriate control groups and, if possible, the recruitment of naïve subjects.

Some of the findings in this review do not support the imperative of specificity, i.e. precise identification of a manipulable lesion and the exact level and side of the manipulation.