This is the first systematic analysis of force-time parameters and electromyographic characteristics associated with two manual HVLA manipulations delivered by a clinician and following one another in quick succession. Our finding that the second thrust occurred with higher peak force and faster rates of force application agrees with the literature that reports experienced clinicians are able to modify these force-time parameters to achieve a “successful” spinal manipulation [20,21,22]. Furthermore, our general finding of greater electromyographic response and shorter delay is also consistent with the literature which reports that with higher peak forces and faster rates of force application (as seen with our second thrusts), neural responses are generally increased [4,5,6, 9,10,11,12,13,14,15,16,17,18,19]. The congruence of our results with the published literature is encouraging, especially considering the significant methodological differences between our work and those used in previous studies (outlined earlier in this manuscript).
However, in contrast to our results, Currie et al. reported that participants who are experiencing low back pain (i.e., are symptomatic) have decreased electromyographic responses and increased electromechanical delays following spinal manipulation of the lumbar region compared to a corresponding asymptomatic group . However, manipulative forces were not directly measured in that study, but were estimated based on a prediction algorithm derived in a different study. Therefore, the precise onset of the treatment thrust (needed to calculate the electromechanical delays) is associated with some uncertainty, and the absolute forces applied during the manipulations cannot be compared with confidence between the asymptomatic and symptomatic groups. For example, it could be that the different electromyographic responses were related to systematic differences between the treatment forces applied to the two experimental groups.
In contrast to this, our study directly measured the forces applied during manipulation and thus the exact timing of thrust onset. Our results support the hypothesis that differences in treatment forces influence electromyographic responses and electromechanical delays. Specifically, our results show that manipulations with a greater rate of force application and more forceful thrusts result in greater electromyographic responses with less electromechanical delay. This was observed in manipulations delivered to the lower cervical and upper thoracic spine in the symptomatic group and for the thoracic region in the asymptomatic group. We also observed this pattern with manipulations delivered to the upper cervical spine in both groups and in the lower cervical spine in the asymptomatic group, but the results did not reach statistical significance, likely due to an insufficient number of comparisons in these regions. Interestingly, there was a greater absolute force delivered at both the upper and lower cervical spines for the asymptomatic population. This systematic difference between the two groups is curious, especially considering that a single practitioner delivered all manipulations. Typically, manipulations delivered by a practitioner to a spinal region (e.g. neck) are consistent  while thrusts delivered to the same region by different clinicians are highly variable . One possible reason for this difference between populations could be previous experience with spinal manipulation. Specifically, in the asymptomatic group, the participants were naïve to spinal manipulation prior to their involvement in the study while the symptomatic group was recruited from the existing patient base of the chiropractor. Thus, factors such as participant familiarity, anxiety related to cervical spine manipulation and the practitioner-patient relationship may have affected the level of force applied by the chiropractor.
Many clinicians judge the success of an HVLA manipulation by the presence of a cracking, clicking or popping sound – commonly known among manual therapists as cavitation [23,24,25]. The first study on cavitation associated with HVLA manipulation was conducted on metacarpophalangeal joints. It has been suggested that an increase in the joint space and an associated increase in joint volume with the HVLA manipulation caused the collapse of intra-articular gas bubbles which were responsible for the sound of cavitation . However, more recent studies using advanced imaging to study both metacarpal joints in the hands and zygapophyseal joint spaces in the cervical spine do not support this earlier hypothesis. Rather, it was reported there was no evidence of gas within the joint space nor an increase in joint space immediately post-manipulation. Further to this, no vacuum phenomena were seen [44, 45]. Thus, the mechanisms underlying spinal joint cavitation are still unknown  and they may differ from those observed in metacarpophalangeal joints. Despite reports that cavitation elicited by HVLA spinal manipulation causes reflex responses  and that spinal manipulations delivered to the low back that elicited cavitation were associated with decreased sensitivity of muscle spindles of paraspinal erector spinae muscles , the current results cannot support this hypothesis. Rather, the current results support the literature arguing that cavitation elicited by HVLA spinal manipulation does not by itself cause reflex responses [29, 30]. Irrespective of this debate, many practitioners will immediately deliver a second thrust in a clinical situation if cavitation is not achieved with the first thrust [23, 26, 27].
As reflex responses associated with spinal manipulation typically occur within 500 ms of the onset of thrust, it is possible that the electromyographic response associated with the second thrust may not be independent of that occurring with the first thrust. Indeed, the delivery of two thrusts within a short period of time made it impossible to define a reliable preload force prior to the second thrust. This is a salient point as modification of preload forces are another force-time parameter important for altering electromyographic responses associated with manipulation [7, 9, 47]. However, inspection of the data suggests that the electromyographic responses elicited by the first thrust returned to baseline prior to the onset of the second thrust. Therefore, there was no direct summation of remnant electromyographic signal from the first thrust with the second thrust. Thus, we believe that it is likely that the same/similar differences would be observed if the two thrusts were measured independently. But there is the distinct possibility that even though the electromyographic signal of the first thrust had subsided prior to the second thrust, the reflex system may have been primed differently for the second compared to the first thrust and thus may have affected the electromyographic response of the second thrust. However, independent of what the exact mechanisms were for the increased reflex responses in the second compared to the first thrust, they would represent what happens in a clinical situation, and potentially might affect treatment outcomes. Further investigation into differences between preload forces and neural responses and the timing of manipulations that are delivered following one another in quick succession should be conducted.
It is possible that the electromyographic responses recorded in this study were inconsistent due to a number of factors. Firstly, variation in electrode placement between subjects may have occurred, resulting in the recording of electromyographic responses from different parts of the same muscle between participants. However, all possible care was taken to ensure that electrode placement was consistent between participants despite differences in body size, shape and anatomy. Furthermore, surface electromyographic is a reliable  and commonly used instrument for the measurement of electromyographic response to spinal manipulation [10, 12,13,14]. Secondly, it is possible that physical differences (e.g., size, weight, somatotype) between participants may have affected the force-time parameters of manipulation such as the line of drive, level of force applied and speed of the thrust. These differences could feasibly have changed the anatomy affected by the thrust and thus the electromyographic responses associated with manipulation. While we attempted to reduce this variability by using the same chiropractor with > 30 years’ experience delivering manual HVLA spinal manipulations in both studies, there were differences between the peak forces, rate of force application and electromyographic delays between the asymptomatic and symptomatic groups in both the first thrust and the second thrust. However, this variability is impossible to control when delivering a manual HVLA spinal manipulation to a human. Furthermore, as we compared responses between two thrusts delivered to the same participant, the effect of size or body type differences would be accounted for in the study design. Further, in some muscles (e.g., left latissimus dorsi, asymptomatic cohort), the electromyographic responses were very small and thus, a small absolute difference between two thrusts would become artificially large when the second thrust was normalized to the first. Thus, we recommend the reader exercises caution when interpreting the percentage difference between the two thrusts.
It is also possible that design limitations of the pressure pad used to record force-time parameters may have influenced our results. Specifically, the pad can only measure forces perpendicular to its surface. Any shear forces are not measured thus the reported forces tend to underestimate the actual forces applied by the chiropractor. Further to this, the pad is flexible and conforms to both the participant’s anatomy (spine and surrounding soft tissues) and the clinician’s hand. Bending of a pressure sensor would give a force that is not applied. However, the pressure pad is constructed with small elements that are spaced apart, and the spacing allows for great bending of the pad without bending of the sensors in the pad. Therefore, we do not believe that bending of sensors caused artifactual forces, and because of the repeated nature of the comparisons with matched contact positions, such artifacts would be expected to be the same for a given comparison pair, thus not affecting differences observed between the first and second thrust. The space between the sensors should not affect the forces measured, as according to Newton’s laws action is equal to reaction, and the empty space between sensors cannot support forces, so all forces applied by the chiropractor are measured by the contact sensors. Finally, we collected force data at a sampling rate of 200 Hz, which is reported to be adequate to describe HVLA manipulative thrusts . Despite this, it is possible the peak forces may be somewhat underestimated. However, we think this underestimation would be relatively small, and it would be a systematic error that would not affect the relative differences between the 1st and 2nd thrust measurements. Nevertheless, the highest rate of force application we measured was 3710 N/s. Sampling at 200 Hz means that we get a force sample every 5 ms. In the worst-case scenario, our data points measured would be as far away as possible from where the actual peak force occurred, which would be exactly between two measured points (or 2.5 ms from the actual force peak). If so, and at the highest rate of 3710 N/s, we would potentially miss the peak force by 3710 N/s · 0.0025 s = 9.3 N. The mean peak forces for these measurements with the highest rate of force applications were 573 N (Table 2 in the manuscript). Therefore, an underestimation of 9.3 N would correspond to an error of about 9.3/573 = 0.016 or about 1.6% of the actual value. Since the rate of force application is always highest somewhere in the middle of the manipulative thrust, and the rate towards the peak force is much smaller (it becomes zero at the peak force), we can safely say that the sampling rate would cause a maximal error in peak force calculation that is less than 1%. Further, this potential offset of treatment forces does not influence the timing of events occurring during the manipulation (e.g., onset of thrust, peak force), which is what were used to analyze the electromyographic responses.
Additionally, the order of the manipulations was non-random – each participant underwent manipulation from C1-T4 in the same order and on the same side. Thus, it is possible that there may have been an order effect present. Specifically, there may have been both ascending and/or descending effects from both the spinal cord and/or brain. However, if this was the case, it is reasonable to assume, although not proven, that it would have affected the symptomatic and asymptomatic population in a similar manner. Regarding a possible order effect between the first and second thrusts, it has been reported that there is an attenuation of the H-reflex response with cervical spine manipulation , suggesting that the electromyographic responses associated with Ia muscle spindle reflexes might be depressed for the second compared to the first thrust. We observed the opposite of what a depressed H-reflex response would suggest: an increased electromyographic responses associated with the second thrust (the onset of which occurred after the electromyographic response returned to baseline following the first thrust). However, since we expect that the reflex responses measured in our study arise from a multitude of sources, and not only the Ia muscle spindle pathway, and since there are no reports on sensitivities of other reflex pathways following cervical spinal manipulation, it is hard to speculate if there was an order effect or if the increased reflex responses in the second thrust were a reflection of the altered mechanics compared to the first thrust. Independent of the answer to this question, it appears that a second thrust given in a clinically relevant situation enhances the reflex response and may have implications for the treatment outcomes.