Key results

We investigated the associations between the onset and disappearance of pain and changes in total, vigorous, moderate, and low PA in men and women based on data from a large, nationwide representative longitudinal study. Our hypothesis of an association between the onset of pain and decreased PA was rejected for men as PA significantly increased with the onset of pain. No statistically significant association was observed for women. Our second hypothesis of an association between the disappearance of pain and increased PA was confirmed for women, but not for men. Our assumptions regarding differences between men and women in the associations between pain and different PA intensities could not be confirmed.

Interpretation

To the best of our knowledge this is the first study investigating different effect sizes of changes in PA in association with the onset and disappearance of pain.

Our finding for women, that PA increased when pain disappeared, was as expected and in line with a study of a Norwegian population, despite methodological differences.8 The Norwegian study had only a time span of 12 months compared to 3 years in our study. Furthermore, the primary analysis in that study was not stratified by sex. By contrast, our finding for men that PA increased when pain started was surprising and not in line with the Norwegian study.8 However, it should be noted that the previous study did not investigate differences in the effect sizes depending on the direction of change in pain. For long-term changes in patients with osteoarthritis, an association between reduced pain and increased PA has been found in another previous study,33 which is again in line with our study results.

Our different findings for men and women match the existing literature. Landmark et al. found a statistically significant interaction between exercise and sex for pain (P < 0.001).8 However, their secondary stratified analysis revealed similar findings for men and women.8 Our analysis used PA as an outcome and our findings are the opposite, that is, we did not find a significant interaction between pain and sex on PA—which might be explained by a lack of power. Future research in this area is therefore required.

To the best of our knowledge this is the first study suggesting that PA intensity plays a role in the association between the onset and disappearance of pain and PA. However, still in line with the literature is our observation that vigorous PA was driving the changes in men while moderate and low PA intensities were the driving factors for changes in PA among women, since these are the preferred PA intensities of the sexes.8, 9

In summary, our findings in conjunction with the existing literature allow the following two possible but contradictory conclusions.

One option would be that there is no association between the onset and disappearance of pain and changes in PA, since some of our analyses did not reveal statistically significant differences, and the ones we found could still be by chance and driven by unobserved time-dependent variables such as IQ or intake of pain medication. This explanation is in line with our observation of no statistically significant differences between men and women; however, this specific observation as well as the (not significant) results from the analysis including only participants experiencing an onset/disappearance of severe pain could also be due to the sample size being too small.

The second option for explaining our findings is the existence of subgroups in the association between pain and PA. This idea is driven by our observation that an association with increased PA was seen for the onset (in men) and the disappearance (in women) of pain. Thus, PA increased independent of the direction of pain (onset vs disappearance). Accordingly, there seem to be some individuals with a positive and some with a negative association between pain and PA. A potential factor moderating the association could be the intensity of the PA. Another factor could be sex; however, it did not moderate the association between pain and PA in our analyses. However, the study from Landmark et al. supports the idea of sex being an important factor in the association. Additional influencing factors could be the origin and location of pain (e.g., acute or chronic, and back or stomach pain) in conjunction with pain-coping strategies. Some patients might have learned that increasing PA is one way to control, reduce, and cope with pain, as suggested, for example, for patients with osteoarthritis,34 while other participants assume all pain to be a warning sign and thus stop any activity. Finally, some participants might be tenacious in the pursuit of their goal, and ignore their pain. However, to the best of our knowledge these factors have not yet been investigated in within-individual association studies. The existence of subgroups would explain our surprising finding that the onset (in men) and disappearance (in women) of pain was associated with increased PA.

Future research should focus on investigations regarding the underlying mechanisms of pain, and on any effect of location, origin, and intensity, on the association between the onset and disappearance of pain and PA.

Strengths and Limitations

One major strength of this study was that longitudinal data were taken from a large representative population-based study, which allowed us to perform stratified asymmetric analyses due to the size. A second strength was the investigation of within-individual associations in pain instead of making between-individual pain comparisons, since pain is a subjective experience, and hardly comparable between individuals. Finally, fixed effect models eliminate the influence of time-constant (both, observed and unobserved) factors on the estimates.

The major limitations of the study were (i) the possibility of reverse causality, and (ii) the tools for measuring our variables of interest. It should be noted that the possibility of reverse causality cannot be dismissed in our study (e.g., from PA to onset of pain). Therefore, future research, for example, based on panel instrumental variable approaches (when valid instruments are present) or based on dynamic panel-data estimations using maximum likelihood and structural equation modelling35, 36 (when the requirements are met such as the presence of at least three waves) is indicated. Existing literature offers both the hypotheses that PA is influencing pain and that pain is influencing PA.6, 13 The findings of the present study are of special importance since they indicate that the direction of pain—onset or disappearance—is of importance for the effect size in the association with PA. It is thus of interest to investigate which factors are influencing the association.

Both our key independent variable and our dependent variables were measured with a single item tool. PA in general is difficult to measure and the preferred ways are objective measures such as accelerometers.37 However, they are more expensive than surveys and therefore less suitable for population-based studies. Furthermore, the IPAQ is suitable for evaluating within-individual changes, according to a systematic literature review.21 The single item used to measure pain is part of the SF-8 survey. As part of a validation of that instrument the question was assessed and found valid.25 Nonetheless, the item provides no information on the origin and location of pain, which is a limitation. Thus, the impact of pain origin and location in this association should be investigated in future studies.