The result of this secondary analysis of data from a large RCT suggests that mild AE after MT may improve the chance to be recovered three months after treatment in men seeking care for non-specific LBP and/or NP.
AE was not a prognostic factor when both genders were analyzed together. This is consistent with results in previous studies, where AE after MT were unrelated to outcome after three months [
14‐
16]. Our results show that AE are common and that most cases are mild, which are in line with the results in previous studies [
6,
7,
9].
Strengths and limitations
Important strengths of this study were the large study population, the careful management of confounders and the high response rate. Potential confounders were identified through theoretical and empirical considerations and were available from the extensive baseline questionnaire. Recovery expectation is a well-known prognostic factor for NP and LBP and was a confounder adjusted for in most of the final analyses [
19,
20]. All study participants were patients seeking care for their complaints, thus they may have higher expectations for recovery than persons who do not seek care. However, this is not a source of bias since the study is etiological and recovery is compared between groups. A limitation is that there could be residual and unmeasured confounding bias of the results, caused by use of medication and from sport or work related traumas or overloads. Further patho-anatomical, neuro-physiological and cognitive-behavioral factors may effect recovery and thus potentially confound the associations.
We used the outcome perceived recovery that is considered to increase the external validity of the results [
21] and that is a reliable assessment of current health status in people with musculoskeletal disorders [
22]. Perceived recovery seems to correlate with changes in pain and disability scores during MT [
23].
Even though data from long-term follow-ups is available in the original trial on which this study is based, we decided not to investigate the long-term effect of AE on recovery. This decision was made based on our hypothesis that it would be improbable that a short time reaction after a treatment would impact the result of the treatment in the long term.
The questionnaire used to measure the exposure AE and the patients rating of bothersomeness from AE has not been formally tested with regard to validity and reliability. This means that there may be a risk for non-differential misclassification of exposure, which may dilute the associations studied
. Since the intervention strategy was to give six treatments within six weeks, the absolute majority of the AE questionnaires were filled in within a week after the treatment session. This means that the risk of misclassification of exposure due to long recall periods is low. AE were categorized into nonexistent, mild or moderate, demonstrating the concept of Carnes et al. [
11] with the same NRS values as a previous study [
24]. Bothersome AEs (NRS > 7) were also considered by Carnes et al. [
11]. However, only a very low proportion of patients (3%) reached up to this level in our study sample thus the group was too small to study separately. These were included in the category “moderate” in the analyses. Further, choosing the highest value from each of the three questionnaires and calculating the mean of the three sessions does not take into account the potential cumulative effect of multiple AE in a single session. This may constitute a limitation in the classification of the exposure. Furthermore, studying the first three sessions combined do not determine if there is a specific effect on any or some of the individual sessions.
The result of this study adds to the knowledge that recovery from pain is a complex concept. Some tissues like intervertebral discs and ligaments, compared to muscles, responds slowly and perhaps incompletely to biomechanical chances [
25]. After biomechanical unloading procedures achieved by MT and conditioning, the response could also be slow. Using MT and/or exercise produces forces interaction between motor and sensory control of the entire spine and related joints. This affects load-sensitive nerve endings located in muscle and tendons providing proprioceptive information including pain [
18]. An alteration in loading of spine from “pain state” to “relief” needs readjustments in sensory-motor control and environment around sensory nerves [
18]. These processes may have a role in the occurrence of AEs. Potential effect mechanisms may not only be biomechanical, but related to context and expectations. This may explain that we did find statistically significant associations in men but not in women.
Treatment related AE may be considered as a response to a biomechanical adaptation where nociceptors are stimulated through the unloading of painful tissues. Emphasis shall however be on recognizing unwanted severe changes so that they can be avoided. Since the presence of mild and moderate AE don’t improve the chance to recover in women, AE should be avoided especially in women.