Our study showed that weight-related factors contribute to the likelihood of Modic changes but not to severe disc degeneration. Whole-body vibration was associated with severe disc degeneration at L5-S1, whereas it was of borderline significance in case of type II Modic changes at L5-S1. To the authors' knowledge, only one population-based study  has previously focused on the determinants of Modic changes. The authors found that heavy physical work for more than 10 years and heavy smoking of more than 20 cigarettes a day were associated with Modic changes .
In a recent population-based study among 40-year-old Danish men and women, the prevalence of Modic changes was 22% . In the present study, we found a higher prevalence of Modic changes (56%), which may be explained at least partly by gender difference. Our occupational cohort consisted only of males and a recent study observed that all Modic changes associated significantly with male gender . The high prevalence of Modic changes in the current study compared to the Danish study may also be due to genetic homogeneity, which has been observed for many monogenic diseases in Finland , or due to the older age of our population. The prevalence of Modic changes has been found to be associated with age [2, 19, 20]. Similarly, we found a close association between age and the prevalence of changes as an evidence of their degenerative etiology. In addition, Modic changes most likely occurred at L4-5 and L5-S1.
In the present study, Modic changes were associated with weight-related factors (BMI, waist circumference). However, these factors were not related to disc degeneration. In another series of patients, Modic changes were associated with increasing weight, but not with BMI . Obesity has been suggested as a risk factor for disc degeneration, but the results remains controversial [7–10].
In this study there was a significant association between lifetime leisure exercise activity and all Modic changes (type I and II combined) at L5-S1 in the multivariate analyses (data not shown). When types I and II were analyzed separately, no significant associations remained (Table 3). There is only limited knowledge about the association of physical activity and Modic changes. One population-based study observed that the prevalence of high-level physical activity at leisure time did not differ between subjects with Modic changes and disc degeneration and those with only disc degeneration . However, in the Danish study occupational physical load was significantly higher in the subgroup with Modic changes and disc degeneration in comparison with disc degeneration but without Modic changes, whereas our study subjects had non-physical jobs. Videman et al.  detected a relation between vigorous exercise and lumbar degenerative changes among former Finnish elite athletes. However, they did not evaluate Modic changes separately. Recent studies on identical twins have shown that 70% of intervertebral disc degeneration can be attributed to familial factors rather than to the mechanical environment . Endplate is, however, a weak link of the spine in compression, and always fails before the intervertebral disc, even if the latter is injured before loading commences . We speculate that mechanical loading damages the endplate and may lead to activation of the degeneration. This may be first visible in MR images as a more active Modic type I lesion, and if mechanical loading continues, it converts to a more stable type II lesion [2, 20].
Smoking has been suspected to carry deleterious effects on the intervertebral discs. According to a systematic review, smoking is associated with LBP . In our study population, there was a trend between smoking (assessed as pack-years) and a higher prevalence of Modic changes (type I and II combined) at L5-S1, but no association with severe disc degeneration. Although an association between smoking and disc degeneration was observed in an earlier study on Finnish twins , the association between smoking and disc degeneration was not confirmed in a later study . The study on effect of smoking on Modic changes should be replicated in another population-based cohort. At the moment, we are unable to either disprove or support the hypothesis that smoking is a risk factor for degenerative imaging findings.
Another factor of environmental exposure, alcohol, has been associated with many organ abnormalities and cognitive deficits. However, little is known about the effects of alcohol on the intervertebral discs. We found no association between alcohol consumption and Modic changes or disc degeneration. It therefore seems likely that alcohol does not have deleterious effects on the intervertebral discs.
Occupation (train engineers vs. sedentary factory work) or the duration of exposure to whole-body vibration was of borderline significance for type II Modic changes located at L5-S1. Interestingly, whole-body vibration was associated with severe disc degeneration, but only at L5-S1. There is evidence of an association between vibration and degenerative spinal changes  but this could not be confirmed in a high-quality twin study with discordant occupational driving exposure . Although our findings indicate that exposure to whole-body vibration may accelerate disc degeneration (and possibly Modic changes), we cannot wholly exclude a chance finding.
Our results on the determinants of Modic changes are relevant as recent publications suggest that Modic changes are associated with low back symptoms [13, 14, 21]. Furthermore, changes located at L5-S1 and Modic type I lesions may be more likely to be associated with pain symptoms than other types of Modic changes or changes located at other lumbar levels . In this context it is interesting that the determinants of Modic changes and severe disc degeneration differ especially at L5-S1 level.
The strength of our study is based on the occupational cohort consisting of train engineers and sedentary factory workers of similar age range. Additionally, all train engineers were full-time train drivers from the northern Finland, which ensured a very similar exposure to whole-body vibration. The factory workers consisted of paper mill and chemical factory workers from the same area with only sedentary job and no vibration exposure.
Limitations of our study are based on its cross-sectional nature, which is a common problem in epidemiological studies. Even if selected determinants can be reasonably well assessed retrospectively, their temporal relation to the development of Modic changes or disc degeneration can not be demonstrated. The effect of age, and maybe of genetic factors, may also dilute the effect of environmental determinants. Additionally, especially the assessment of lifetime leisure exercise is susceptible to recall bias. The validity of the original questionnaire has been well documented with respect to cardiovascular outcomes . However, it estimates caloric expenditure more than weight bearing or strength-related dimensions of physical activity. Thus, its sensitivity to detect dimensions of physical activity may not be optimal. Furthermore, we did not have precise measurements of whole-body vibration exposure. When studying lifetime effects, such precision is not feasible due to variations in train models and railway conditions.
Based on the findings of our study, the determinants of Modic type I and II changes are different. The determinants of Type I remain unidentified, whereas age and weight-related factors (BMI, waist circumference) were related to type II changes. This may be partly explained by the low number of subjects with type I changes. However, type I changes are also associated with an ongoing degenerative process which over time converts to Type II or normal bone marrow . Finally, the influence of determinants may also vary at different stages of the degenerative process.