‘Correlation doesn’t imply causation.’ It’s a commonly expressed caution in the health sciences. What it means is that things can tend to go together without necessarily causing each other. The classic example that statistic professors like to give is that air conditioner use is significantly correlated with street crime. Does the use of air conditioners cause street crime? No, of course, not and yet they do tend to go together. It’s actually only because they both tend to occur in the summer. When it’s hot outside, it’s true that people tend to use their air conditioners and it’s also true that people tend to loiter outside in the city, getting into trouble more often than in the cold winter months when people tend to stay indoors. With the example, we can see that just because things tend to go together they don’t also always cause each other.
What about things that hardly go together or things that don’t go together at all? Could we say that low levels of correlation or an altogether lack of correlation imply causation? On the face of it, it seems absurd to think that lack of correlation or even minimal correlation might imply causation. Who would think that A causes B when A and B have no relationship to one another or only a minimal association with one another? That inference, however, is exactly what we assume when we think that degenerative disc disease causes chronic back or neck pain.
The logic of correlational studies of degenerative disc disease and chronic spine pain
Degenerative disc disease is a common explanation for chronic back or neck pain. Degenerative disc disease is a general phrase that refers to a number of conditions or changes of the spine, such as loss of disc height, disc bulges, annular tears, disc herniations, endplate changes, neuroforaminal and spinal canal stenosis, among others. Healthcare providers typically identify their occurrence with either X-ray, CT or MRI scans. When these tests identify degenerative disc disease in the spine of someone experiencing chronic back or neck pain, many healthcare providers and their patients consider that the degenerative disc disease is the cause of the chronic pain. In other words, they believe that the degenerative disc disease is painful.
Healthcare providers and their patients commonly justify particular types of treatment with the notion that degenerative disc disease is causing chronic back or neck pain. Spine surgery and interventional pain procedures, in particular, but also certain types of physical therapy, all attempt to reduce pain by modifying specific types of degenerative changes of the spine. Again, the underlying justifying belief is that such degenerative changes of the spine are painful.
How valid is this justification? In other words, how true is it that degenerative disc disease causes chronic back or neck pain?
The question is actually tricky to answer, especially if we want to show that it is in fact true. Theoretically, the best way to answer it would be to run a true experiment, where we induce various degenerative changes to the spine in a number of people and see whether they have pain in the back or neck (wherever the degenerative change was induced). However, we don’t know how to make degenerative changes to the spine, at least not exactly as they occur naturally. Additionally, even if we did know how, there would be all sorts of ethical problems in producing degenerative disc disease in human subjects. For instance, we don’t know how to reverse it and so we’d be inducing permanent damage to people. Because of these challenges, we don’t tend to perform true experiments when looking at the relationship between degenerative disc disease and chronic neck and back pain.
Instead, we generally rely on a different form of research in the scientific study of chronic pain – correlational research. In correlational research, we measure certain variables, such as the occurrence of degenerative disc disease and back or neck pain, and see whether they tend to go together. If two things, such as degenerative disc disease and pain, are highly correlated (i.e., they tend to go together almost always), we can at least say that one may cause the other. As we discussed in the introduction of this piece, though, we cannot say with confidence that there is a causal relationship when two things are highly correlated. There may be other factors that cause the two to occur together all the time (i.e., think of the air conditioner use and street crime example). Nonetheless, it is helpful to see whether they correlate together and what the strength of the correlation is. The reason is that if they don’t correlate at all or if the correlation is very weak, then we can say with confidence that they don’t cause each other.
To understand, we need to look at the following reasoning: when one thing causes another thing, they must occur together in some demonstrable way; if, however, when one thing occurs, another thing may or may not occur, they have no relationship to each other and so we cannot say that there is a causal relationship. A causal relationship presupposes a correlational relationship. Now, as we have said before, a correlational relationship is not enough to demonstrate a causal relationship, but to have a causal relationship, there at least has to be a correlational relationship. We have to at least be in the right ballpark, as it were. Without a correlational relationship, though, we can say with much greater confidence that we are not even in the right (i.e., causal) ballpark. In other words, if we find no correlation, we can safely assert that there is no causal relationship.
A similar, albeit slightly different logic holds with weak correlations – the situation in which two things go together in some minimal ways. With such correlations, we can say that there may be a causal relationship between the variables, just as we said with strong correlations, but we can’t know for sure because correlation doesn’t imply causation. Now, with weak correlations, however, we can go a step further. We can assume that even if there was a causal relationship between the two variables, we know that there has to be more to the picture than simply the two weakly correlated variables. In other words, there has to be additional causal variables coming into play, because if there weren’t the correlation would have to be strong. The weak correlation, even in cases where we might assume a causal relationship, shows that the variable is only part of the cause – and only a small part at that.
It would be helpful to take an example. Let’s suppose that we did a study of the relationship between the presence of eating utensils and cookware in the kitchen on the one hand and how much food people ate on the other hand. In a sample of 1,000 people, we measured a) the extent to which they had eating utensils and cookware in the kitchen and b) how much they ate on average over the course of a month. Let’s further suppose we found a weak, statistically significant, relationship between our two variables of interest. It’s a fantastical example, of course, but we might see how it could be true: the presence of eating utensils and cookware could play some small role in how much you ate on average – if you don’t have a way to prepare and eat food, it could affect how much you eat. However, we can also immediately see that there’s more to the picture in terms of what goes into the fact of how much food people eat. The extent to which you have ways to prepare and eat food are not the only variables that can lead to eating. People can eat raw foods; they can eat with their hands; they can buy prepared foods in the grocery store, delis, and fast food restaurants; they can go out to eat in sit-down restaurants; and so on. Personal characteristics of the people can also play a role: how hungry they are or how much stress they are under or how busy they are can also affect how much people eat – even in people who don’t have adequate means to prepare food. In all these ways, we can see that a weak, statistically significant, correlational relationship cannot explain the whole nature of the relationship between two variables, even when we assume that they are in some ways causally related.
As we will see in the following review of the correlational research on degenerative disc disease and chronic back or neck pain, degenerative changes of the spine fit into one of these two categories: they either have no relationship at all with chronic back or neck pain, or they are only weakly related to chronic back or neck pain. As such, we can conclude with confidence one of two things, depending on the type of degenerative disc disease we are discussing. First, in the case where research repetitively shows no correlational relationship between certain types of degenerative disc disease and chronic back or neck pain, these particular types of degenerative changes do not cause pain, despite the common belief that they do. Second, in the case where research repetitively shows a weak, statistically significant, correlational relationship between certain types of degenerative disc disease and chronic back or neck pain, these particular types of degenerative disc disease may play some role in producing pain, but we know that it is only a minimal role, even if we assume that the correlation reflects a causal relationship. In other words, the weak correlation between certain types of degenerative disc disease and pain shows that the lion’s share of what’s causing the pain is something else entirely. This statement too stands in stark contrast to the common belief that degenerative disc disease is the predominant cause of chronic back or neck pain.
Let’s, then, review the correlational research on the relationship between the different types of degenerative disc disease and chronic back or neck pain.
Correlation (or lack thereof) between pain and degenerative disc disease
In a review of early studies, van Tulder, at al., (1997) found weak significant associations between back pain and disc space narrowing, osteophytes, and sclerosis, with odds ratios in the range of 1.2-3.3. Other degenerative changes, such as spondylosis, spondyolisthesis, and kyphosis had no relationship to back pain.
In their review of the literature on the natural history and clinical significance of disc herniation, Grande, Maus, and Carrino (2012) conclude that there is no relationship between any characteristics of disc herniation, including size or severity, and subsequent symptoms of patients.
Mitra, Cassar-Pullicino and McCall (2004) found no relationship between evidence of an annular tear in the disc and pain.
Jarvinen, et al., (2015) found no significant correlation between Modic 1 or Modic 2 changes and low back pain.
de Schepper, et al., (2010) studied the relationship between osteophytes, disc space narrowing, and low back pain. They found that disc space narrowing, especially, at more than one level, was most significantly related to low back pain, but only weakly, with an odds ratio of 2.4.
In a more statistically oriented review, Chou, et al, (2011), systematically searched the literature and combined studies to determine the odds ratio for having lumbar degenerative changes and chronic low back pain. They found a significant, yet weak, association between the two. The range for the odds ratio was between 1.8-2.8.
Livshits, et al., (2011) found a significant relationship between all degenerative changes of the spine and pain with an odds ratio of 3.2.
In a study published after the Chou, et al. findings, Nemoto, et al., (2012) found a significant correlation between vertebral osteophytes and low back pain, but the odds ratio was a little greater at 3. In contrast to the de Schepper, et al, study cited above, they found no correlation between disc space narrowing and back pain.
All these data are what likely led Bogduk (2012), one of the founding fathers of interventional pain management, to conclude, “Degenerative changes [of the spine] lack any significant correlation with spinal pain.”
These findings are also similar to the relationship between degenerative disc disease and pain related disability. Quack, et al., (2007) found either no relationship or only weak correlations between lumbar degenerative changes and mobility. Sirvanci, et al., (2008), found no significant relationship between lumbar spinal stenosis and perceived disability, as measured by the Oswestry Disability Index. Lohman, et al., (2006) found no relationship between spinal stenosis and pain or scores on the Oswestry Disability Index. Remes, et al., (2005) in a cohort of patients who underwent fusion for spondylolisthesis twenty years ago, found no relationship between lumbar degenerative changes and the same measure of disability. Similarly, looking at a number of biological and lifestyle factors, Wilkens, et al., (2013) found that degenerative changes as found on imaging failed to correlate with perceived disability one year later as measured by the Roland-Morris Disability Questionnaire.
This review of the literature shows that the various types of degenerative disc disease either have no relationship to pain and disability or only a weak correlation to pain and disability. What this means is that degenerative disc disease is likely not painful. At best, it plays a minimal role in the cause of chronic back or neck pain. The true cause of chronic back or neck pain must be something else entirely.
Bogduk, N. (2012). Degenerative joint disease of the spine. Radiology Clinics of North America, 50(4), 613-628. doi: 10.1016/j.rcl.2012.04.012
Chou, D., Semartzis, D., Bellabarba, C., Patel, A., Luk, K., Kisser, J. M., & Skelly, A. C. (2011). Degenerative magnetic resonance imaging changes in patients with chronic low back pain: A systematic review. Spine, 36, S43-S53. doi: 10.1097/BRS.0b013e31822ef700
Del Grande, F., Maus, T. P., & Carrino, J. A. (2012). Imaging the intervetebral disk: Age-related changes, herniation, and radicular pain. Radiology Clinics of North America, 50(4), 629-649. doi: 10.1016/j.rcl.2012.04.012
de Schepper, E., Damen, J., van Meurs, J. B., Ginai, A. Z., Popham, M., Hofman, A., Koes, B. W., & Bierma-Zeinstra, S. M. (2010). The association between lumbar disc degeneration and low back pain: The influence of age, gender, and individual radiographic features. Spine, 25(5), 531-536. doi: 10.1097/BRS.0b013e3181aa5b33
Jarvinen, J., Karppinen, J., Niinimaki, J., Haapea, M., Gronblad, M., Luoma, K., & Rinne, E. (2015). Associations between changes in lumbar Modic changes and low back symptoms over a two year period. BMC Musculoskeletal Disorder, 16, 98. doi: 10.1186/s12891-015-0540-3
Livshits, G., Popham, M., Malkin I., Sambrook, P. M., MacGregor, A. J., Spector, T., & Williams, F. M. (2011). Lumbar disc degeneration and genetic risk factors are the main risk factors for low back pain in women: The UK twin spine study. Annals of Rheumatic Disease, 70(10), 1740-1745. doi: 10.1136/ard.2010.137186
Lohman, C. M., Tallroth, K., Kettunen, J. A., & Lindgren, K. (2006). Comparison of radiologic signs and clinic symptoms of spinal stenosis. Spine, 31(16), 1834-1840.
Maus, T. (2010). Imaging the back pain patient. Archives of Physical Medicine and Rehabilitation, 21(4), 725-766. doi: 10.1016/j.pmr.2010.07.004
Mitra, D., Cassar-Pullicino, V. N., & McCall, I. W. (2004). Longitudinal study of high intensity zones on MR of lumbar intervetebral discs. Clinical Radiology, 59(11), 1002-1008.
Nemoto, O., Kitada, A., Naitou, S., Tsuda, Y., Matsukawa, K., & Ukegawa, Y. (2012). A longitudinal study for incidence of low back pain and radiological changes of lumbar spine in asymptomatic Japanese military young adults. European Spine Journal, 22, 453-458. doi: 10.1007/s00586-012-2488-4
Quack, C., Schenk, P., Laeubil, T., Spillmann, S., Hodler, J., Michel, B. A., & Klipstein, A. (2007). Do MRI findings correlate with mobility tests? AN explorative analysis of the test validity with regard to structure. European Spine Journal, 16(6), 803-812.
Remes, V. M., Lamberg, T. S., Tervahartiala, P. O., Helenius, I. J., Osterman, K., Schlenzka, D., Yrjonen, T., Seitsalo, S., & Poussa, M. S. (2005). No correlation patient outcome and MRI findings 21 years after posterior or posterolateral fusion for isthmic spondylolisthesis in children and adolescents. European Spine Journal, 14(9), 833-842.
Sirvanci, M., Bhatia, M., Ganiyusufoglu, K. A., Duran, C., Tezer, M., Ozturk, C., Aydogan, M., & Hamzaoglu, A. (2008). Degenerative lumbar spinal stenosis: Correlation with Oswestry Disability Index and MR imaging. European Spine Journal, 17(5), 679-685. doi: 10.1007/s00586-008-0646-5
van Tulder, M. W., Assendelft, W. J., Koes, B. W., & Bouter, L. M. (1997). Spine, 22(4), 427-434.
Wilkens, P., Scheel, I. B., Grundes, O., Hellum, C., & Storheim, K. (2013). Prognostic factors of prolonged disability in patients with chronic low back pain and lumbar degeneration in primary care: A cohort study. Spine, 38(1), 65-74. doi: 10.1097/BRS.0b013e318263bb7b
Author: Murray J. McAllister, PsyD
Date of last modification: 5-31-2015