Of shiny pictures and poorer outcomes: Spinal MRI and back pain

Diagnosing low back pain is a nightmare. It established that apart from the 15% of back pain cases which can be attributed to a specific spinal pathology, the majority of cases fall under the unsatisfactory umbrella label of “non-specific low back pain”.

I was discussing with a colleague a new review that, while admittedly light on data, suggests that diagnostic tests offer little in the way of reassuring patients. This chimed with my (unreliable) anecdotal clinical experience with back pain wherein the results of X-Rays, MRI scans and the like often seem to strike fear into the hearts of patients and act as obstacles to resuming normal behaviour. In fairness to the technology it’s not really the scans that are the problem, it’s the way that they are (mis)used by clinicians. Powerful images of bulging discs, degenerating joints, partial dislocations or instability are evoked to help explain the patient’s symptoms, with the result that patients might be left with the plausible idea that their spine is what is technically referred to as “structurally buggered” (or “struggered”). Like Lorimer’s magic disc-popping action figure.

We’ve discussed the problems with this before. The best evidence strongly indicates that these stuctural findings on X-Ray and MRI are not clearly related to the onset, severity, duration or prognosis of low back pain. The presence of degenerative changes (see also here), disc pathology, muscle wasting, even spondylolisthesis and spondyolysis are common in those without pain and are poorly correlated with the signs and symptoms of low back pain. It’s counterintuitive but there you are. It is what it is. Be honest though, if you are a clinician – it’s hard not to blame that lack of recovery on “wear and tear” isn’t it? Neat, tidy but unsubstantiated  explanations are hard to shake off.

Recent clinical guidelines reflect this evidence by not recommending spinal imaging in the absence of clear indicators of specific pathology (or “red flags”). But beyond simply not telling us much a paper from last year suggests that MRI scans might achieve something worse.

These researchers analysed a large workers compensation database for back pain claims in the USA. They looked at the relationship between the early use of MRI imaging in people with work-related back pain and clinical outcomes. With this kind of retrospective data it is always difficult to attribute cause and effect. For example MRI might be associated with poor outcome because those with more severe injuries are more likely to get an MRI and at the same time are more likely to experience ongoing symptoms. To try to work around this the authors did a nifty trick by identifying first what factors actually increased the likelihood of somebody receiving an early MRI in their dataset and then dividing the groups into “high propensity and low propensity for MRI”. They then analysed the effect that having an MRI versus not having it had on outcomes in each of these groups while statistically controlling for things like symptom severity, age etc.  If MRI is itself an independent risk factor for poor outcome then we might expect to see it have an effect in the group that did not have those factors but that nonetheless still received an MRI. Like I said: nifty.

And that is precisely what they demonstrated. With big brass bells on. The statistically adjusted results indicate that folk in the MRI group came off of disability 200% slower (from a raw average of 134 days with MRI to 23 days without!) than those who did not have an MRI scan. Perhaps more worryingly while those in the no-MRI groups had a surgery rate of less than 10%, the MRI groups had surgery rates of 80-100%. To be clear in the group who had clinical characteristics that made them less likely to be offered an MRI, but who still received one, 100% underwent spinal surgery. These, you might notice, are large effect sizes.

How do we interpret these findings? Well maybe the results could be explained by the deleterious effects of spinal surgery or maybe they are due to the social and psychological impact of being given a structural diagnostic label that is likely to be spurious in the majority of cases. Perhaps there are influences in there specific to the opaque dynamics of a workers compensation system, with all of its competing agendas.  Most likely it is an murky combination of all of the above and more.

What the study does tell us is that as well as not being very informative, spinal imaging in this group was related to important differences in the treatment paths of these patients and also to a large and unhelpful change in clinical outcome. It’s time to listen to those clinical guidelines and be careful which spines we take pictures of.

About Neil

Neil O’Connell is a researcher in the Centre for Research in Rehabilitation, Brunel University, West London, UK. He divides his time between research and training new physiotherapists and previously worked extensively as a musculoskeletal physiotherapist. He also tweets! @NeilOConnell

Neil is currently fighting his way through a PhD investigating chronic low back pain and cortically directed treatment approaches. He is particularly interested in low back pain, pain generally and the rigorous testing of treatments. He also tends to get all geeky over controlled trials.


van Ravesteijn H, van Dijk I, Darmon D, van de Laar F, Lucassen P, Hartman TO, van Weel C, & Speckens A (2011). The reassuring value of diagnostic tests: A systematic review. Patient education and counseling PMID: 21382687

Carragee EJ, Alamin TF, Miller JL, & Carragee JM (2005). Discographic, MRI and psychosocial determinants of low back pain disability and remission: a prospective study in subjects with benign persistent back pain. The spine journal : official journal of the North American Spine Society, 5 (1), 24-35 PMID: 15653082

Kalichman L, Kim DH, Li L, Guermazi A, & Hunter DJ (2010). Computed tomography-evaluated features of spinal degeneration: prevalence, intercorrelation, and association with self-reported low back pain. The spine journal : official journal of the North American Spine Society, 10 (3), 200-8 PMID: 20006557

Kalichman L, Li L, Kim DH, Guermazi A, Berkin V, O’Donnell CJ, Hoffmann U, Cole R, & Hunter DJ (2008). Facet joint osteoarthritis and low back pain in the community-based population. Spine, 33 (23), 2560-5 PMID: 18923337

Jarvik JJ, Hollingworth W, Heagerty P, Haynor DR, & Deyo RA (2001). The Longitudinal Assessment of Imaging and Disability of the Back (LAIDBack) Study: baseline data. Spine, 26 (10), 1158-66 PMID: 11413431

Kalichman L, Hodges P, Li L, Guermazi A, & Hunter DJ (2010). Changes in paraspinal muscles and their association with low back pain and spinal degeneration: CT study. European spine journal : official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society, 19 (7), 1136-44 PMID: 20033739

Kalichman L, Kim DH, Li L, Guermazi A, Berkin V, & Hunter DJ (2009). Spondylolysis and spondylolisthesis: prevalence and association with low back pain in the adult community-based population. Spine, 34 (2), 199-205 PMID: 19139672

Webster BS, & Cifuentes M (2010). Relationship of early magnetic resonance imaging for work-related acute low back pain with disability and medical utilization outcomes. Journal of occupational and environmental medicine / American College of Occupational and Environmental Medicine, 52 (9), 900-7 PMID: 20798647


  1. Well rounded review! This is why we set up our spinal assessment clinic to reduce unnecessary referral to consultants and further investigations like MRI etc. working with pain management in parallel clinics as well as direct access to back programme and exercises referral scheme we’ve seen a huge improvement in overall outcomes via the Keele start back Ax tool and Oswestry.
    One question on these two outcome measures and their overall validity for a service like this- could I pick your brains one day please?
    Thanks 🙂

  2. Abnormalities found on an MRI scan for the neck and back do not lead to pain. The medical community has chronic pain all wrong. I’m pointing out here that chronic pain does not have physical or structural reasons. It’s psychological and emotional. Here me out now. It’s called TMS. Tension Myositis Syndrome. Also, referred to as Mind Body Syndrome, PPD, etc. There’s a ton of information supporting this condition. Here’s some reasons TMS has not become more popular: doctors are not taught this in medical school, the medical community makes huge money from chronic pain, most doctors don’t want to deal with the psychological aspect(get them in and out of the office as quick as possible), most patients want a drug or something easy to get rid of the pain. The research is available on this subject. You should check out the TRUTH. Most people just don’t want to believe that their mind is the reason for their pain. What causes chronic pain? It’s called TMS. 500 years ago the earth was believed to be flat. Is your world flat or round?

  3. I was wondering if a study has been done like this: Surgeons (i.e. the real “decision makers”/influencers” not the radiologist) look at MRIs of 100 patients (without seeing the patient): 50 of whom have severe, disabling pain and 50 who have no symptoms at all. Can surgeons reliably seperate the two groups? If so, it would show the MRI would be a useful test for classifying severe back pain.

    Neil O'Connell Reply:

    Hi richard,]

    Good question and a nifty idea. I don’t believe precisely that study has been done but the best (imperfect) data we have to date suggests that for most back pain presentations MRI offers little to nothing in terms of diagnostic or predictive ability. So if we removed clear cases of radiculopathy and red flags I would predict that they would not.

    After all if a systematic approach to identifying these abnormalities on MRI does not predict outcome, or indeed the opresence of pain, what special undisclosed wisdom might those surgeons be relying on?

  4. Great article and for what it is worth my ” (unreliable) anecdotal clinical experience with back pain ” also chimes.
    Used to be on an MDT for spinal pain dominated by spinal surgeons with trial my MRI – shame.

    So glad I have bumped into your blog. Thanks

  5. Thanks Neil. This is a topic near and dear to my heart. I hope the nonprofit I work with will be a small part of helping to turn this tide on letting tests and surgeries be god in determining whether we will lead productive, full lives or be a imprisioned by a diagnosis. Thank you for making this drop in the pool of change.

    Neil O'Connell Reply:

    Thanks a lot Cynthia,

    It is worth noting that invalidation of one clinical proces does not necessarily validate another. So for things such as Feldenkrais and other things I noted from your website rigorous evidence is also required.

    Cynthia Allen Reply:

    Skillfully said and I agree. This is why research is part of our mission. Still in the absence of research or even when it is present with clear “evidence” (because we know it is rare that research proves anything for all people and all situations for all times) it is the job of the health care professional or in my case the somatic educator to be present to each person and their unique situation and look for best pathways. The lack of evidence to date should not immobilize us into doing nothing or into doing the same old tired thing when it’s not working or even injuring people.

    This approach maybe significantly easier for me say in my private practice because I see almost entirely people who feel failed by the mainstream approach and most often have had surgery with poor outcomes. However if these same people had seen a physical therapist or pcp or even orthopedist who had, as you have suggested educated the person about unknowns and knowns with back pain and the limitations of MRIs and perhaps even the negative consequences of following that path, perhaps different decisions would have been made. Perhaps PT could have been successful. Perhaps the person would have had more patience with the back pain and confidence in their ability to get better instead of needing a test or surgery. That too would be a good study.

    Anyway keep up the good work.

  6. Hi Neil.

    Here you have also a couple of interesting references.



    May be still good to take some pictures for 10-15% of LBP sufferers, but may be controversial their use for the other 85%. The big bulk are the very big problem.

    And as far as we know, imaging is not improving our diagnosis, therefore, nor the treatment.

    I’m not going to talk about subgrouping and all that stuff. It would be a waste of time.

    But I really would like to state, that if we “follow” the guidelines, we won’t be able to get out from the mud.

    The worlwide accepted diagnostic triage is lacking any scientific evidence, since it’s just a consensus among some “experts” or “authorities”, and it was the best way of classifying LBP. According to them.

    But, as we know, that triage still doesn’t give us any better approach.

    And as I told you in an ealier post, this big group (the non-specific), is a big one, but the sample is heterogeneous.

    How can we compare different treatments with such heterogeneous sample? The results will be as heterogeneous as the sample. So the results and conclusions won’t be valid.

    So we need to encourage for a shift in the direction of the diagnosis. And of course the evidence, regarding the lack of effectiveness of the imaging, is mounting.

    The review I referenced above, concluded “to refrain” from imaging because it doesn’t improve clinical outcomes.

    So we can talk about better treatments but it would be nonsense, since the actual problem is the lack of an appropriate diagnosis.

    If a patient is very keen for a CT or MRI, we as professionals should explain to them why not.

    This is a difficult matter, I know, but if we had all the information, we would be able to give the information to the patient. But the patients is aking for imaging because they “believe” it’s a diagnostic tool, the best. But the evidence, apparently, says otherwise.

    So we have to start (may be), to “educate” patients rather than feeding their believes, believes that won’t give any extra positive information, but could be very negative, because if they learn they have a “slipping disc”, “degenerative changes”, and we don’t explain to them the meaning of these technical words, plus the clinical relevance, we will be strengthening their believes, and that will become a barrier for a recovery. That’s it we will making chronic pain patients.

    Guidelines and diagnostic labels (not only for LBP but for any other pathology) if useless can be a huge barrier for treating patients. But it seems noone is listening.

    We’re learning about mirror neurons, shared pain, the power of words/names, and so on, amazing breakthrough in such difficult field. But we are incapable of just trying to improve the poor approach to LBP, and being “happy” to be under the umbrella of useless labels.

    Neil, I know this is a topic difficult to dicuss, but over years I’ve been very involved in the LBP field, and what I’m 100% is that guidelines are themselves an enemy rather than a friend. What I’m exposing here is a mixture of objective assessment, and from that assessment I tell my opinion.

    I’m not a stakeholder of any approach, but I’m a defender of things well done, and in the LBP, despite all the evidence against the established criteria, we’re still worshippers.

    I don’t expect a change in the following years, not even in this century, but at least I’m happy to read, share and learn from posts like your and other’s.

    Thanks for your time, and good stuff!

    Neil O'Connell Reply:

    Thanks arco,

    I agree that we have a long way to go in finding good answers to low back pain. As we have discussed before NSLBP is entirely unsatisfactory as an explanatory label but we don’t really have anytrhing better at the moment to guide research or practice.

    arco Reply:

    Cheers Neil!

    In 1998 the world leading authorities (Borkan et al and the Cochrane group), said that subgrouping was a way of improving the approach to NSLBP. So the same people that did the Triage, were then saying it was not enough, and from the research point of view they encouraged to perform more RCT’s to find the better treatment for subgroups.

    So o improve the diagnosis to improve our treatments.

    So yes we have another different way, but the question is: “Is anyone listening out there?” Unfortunately, lots of subrgouping approaches to define the same thing, and no firm conclusions, but there is enough evidence in the literature to suggest that may be there is a way, but “Is there a will?”

    Other problems rather than not having other ways of approaching NSLBP are the issues. If we were all going in the same direction, it would be another story.
    Anyway, I don’t want to waste anyone’s time, just sharing my thoughts.

    Thanks for replying.


  7. Hi, good stuff! We all know about the powerful effect of placebo, and here’s a great example of the nocebo effect, clearly also powerful, probably because it’s a big, noisy, important looking machine – I’m sure it would it be less so if it was a small hand held thingy, like “Bones” had in Star Trek? And yes, they are dangerous, people ( few admittedly) have died as a result of MRIs attracting ferous objects – such as oxygen bottles accidentally trolleyed into the room flying across to the magnet, and a CT Scan gives you the equivalent of an extra 5 years background radiation- with an increased risk of some malignancies.
    So, avoid an MRI/CT except in the very narrow group where it is clinically indicated – for what? Conservative management of some previously operated on conditions is just as good- and better in that you avoid the side effects of surgery, including psychological and social.
    If someone else ( not me of course!) has done an MRI how do you avoid the “threatening illness information” then? “those bits are your internal wrinkles & wrinkles don’t hurt” seems to help a few, “good news, you don’t have cancer. infection or arthritis” might help..but might frighten other people- they don’t always hear the “don’t” . I struggle with the right way to say those things! And then the patient comes away from you with a “non diagnosis” of “non-specifc low back pain”, and their work mates say “well why are you still sore, and why can’t you work then?”
    As you say, there are some strongly held public notions that need to be adressed . How? pain is low on the agenda in pretty much every country- we need some sore politicians? But they are likely to be active copers and therefore would not see or understand the issues, and telling them that 20% of the population has long term pain is just going to frighten them off!
    Thanks for the article.

  8. Hi Neil
    many thanks for the stimulating post again – a subject which triggers my somatic markers!

    This is the perennial issue of when to investigate and when not. Given the poor correlation between radiological findings, symptoms and prognosis we are usually left with the scenario clinically of identifying those patients with clear cut surgical indicators (which fail to respond to a trial intervention of 3-4 treatment’s or are too acute to even attempt this). This takes care of perhaps 10 to 15% of the patient cohort which might be good for them but not much consolation for the other 85%!

    The last review I saw in spine (admittedly a few years ago now) indicated a surgical preference for lumbar discectomy about 10 times greater in the U.S compared to Europe. As I recall the long-term follow up at 1, 5 and 10 years showed no significant differences in outcome but the surgical intervention group had a considerably shortened period of acute disability.

    Whilst I’m no big fan of surgery my anecdotal clinical experience suggests that the caliber of the surgeon, the patient selection and compliance are important determinants in outcome.

    At the risk of “lowering the tone” in this distinguished forum I’m going to raise some practical clinical issues which I think need Consideration.

    1. For patient’s who feel a compulsion to seek radiological investigation denying them may become a roadblock to patient management. This may be acceptable in a public health system where policy can be attributed to “remote” guideline committees etc. In private sector healthcare delivery this can be distorted into a perception of clinician incompetence, lack of empathy or simply a credibility issue.

    2 I think the clinical preamble into recommending radiological investigations is of critical importance. Here I think the treating clinician needs to take a calculated risk by predicting the likely findings of scans, x-rays etc. and also by discussing how these likely findings will or will not alter the management strategy. This requires the clinician to put their “head on the bock” and take responsibility.
    Although I don’t know if this has been formally studied, but in my experience it enhances clinician credibility and provides a better foundation for delivering (or withdrawing) treatment on the basis of the efficacy of intervention.

    3. Whilst it is clear that the correlation between radiology, pathology and functional discrepancy remains inconsistent I think it would also be fair to say that observable radiological change represents a relatively well established stage of a process (whether it be degenerative, age related or something sinister). I don’t think we’re at the stage of quantifying the early biological, microstructural or tissue sensitivity (neurobiological) characteristics which are associated with pain but not visible radiologically, or with any other instrumentation for that matter.

    So in conclusion I would wholeheartedly endorse your recommendation for exercising caution in this regard but would also make the case for the clinician making a conscious decision of when to investigate or not – not just on clinical criteria but by integrating elements of the psychosocial framework which evaluate the relevance(psychological )for the patient and enhance the therapist patient relationship in order to optimize the compliance with the management program.

    Here are some further clinical discussions for those interested:



    Neil O'Connell Reply:

    Thanks David, some very interesting arguments.

    Points 1 & 2: The suggestion that we might scan some folk in the 85% because they wish us to and because as a skilled clinician we can intuit that it would help them to make progress is an interesting one. Particularly if we accept that such scans will tell us nothing useful about the condition that we couldn’t tell through clinical examination alone. I agree that if you are going to do that, the way that you frame the reasons and context for that investigation with the patient would be vital. However I might also venture that this skilled clinician might avoid what could be described as an act of benign fraud, and better use their communication skills to persuade the patient of the many good reasons that such a test is unnecessary. I understand how private practrice influences these decisions in the real world but ultimately that is something of a corruption of the clinical process to a different bottom line.

    Point 3. I think a better way to describe the correlation from the better studies would be “not there” rather than “inconsistent”. In a way that is a bit like when I hear people talk about a treatment having “not enough evidence for it”, when in fact there exists a consistent body of evidence against it.

    If these structural changes were key players in the clinical presentation of back pain then a logical prediction would be that we would consistently see these relationships in the data. But we don’t. So I am not sure that this well established process is of much clinical importance.

    The point you make about local biochemical changes is a good one. I fully agree that just because our best imaging technology can’t identify a key player in the tissues of the spine it doesn’t logically follow that there isn’t one. We simply don’t know. But the point on imaging remains – if it tells us nothing and if there is a measurable risk associated with it then I would still suggeast that in those 85% we should not do it.

    David FitzGerald Reply:

    Fair point Neil “not there v’s inconsistent”!

    I guess it comes down to a judgement call on how much value a patient places on specific radiology to explore their issue.

    I generally try to go with the idea of the scan unlikely to influence the management strategy. I Put in a category of “nice to have for exclusion” but not clinically relevant.

    At the risk of being vulgar for those who’s health insurance cover’s this cost without question or those who expect investigations because they pay their taxes the challenge becomes more tricky – but still not to be avoided.

    I must confess I don’t think it’s too hard to select patients where scan’s display the suspected pathology – so I don’t think there is a big risk there.
    But as for the other 85%!

    I like Peter O’Sullivan classification system on the basis of a unifying approach, allbeit lacking detail on cognitive input.

    If memory serves me wasn’t there also a study from the Netherlands comparing the type of educational information patients recieve about their condition the more graphic being worse.

    Mike Sullivan also studied clinicians attitudes which were significant determinants on outcome.

    We really do have to watch our P’s and Q’s!!


  9. Mark Brennan says

    Hi Neil
    thanks for bringing this to our attention. It certainly does back up the approach of the seemingly limited use of MRI in low back pain. However your point early on is that it not so much the scans as to what the interpretation and the explanation given to the patients about the scans that potentially lead to worse outcomes. I imagine this was not highlighted in the study. Maybe looking at early MRI with a view to specific education about scans and what are “normal findings” with a pain education bias would achieve, maybe lots or reassurance. Many patients often feel the scan is the be all and end all. Perhaps a scan with the right explanation from the clinician may bring about a more favourable outcome as the patient feels reassured that nothing is being missed as they have had a scan but, crucially, with a good explanation. Also I think we are in danger of righting off scans all together, they do after all provide information when used correctly. Patients who don’t always present with the 15% classic spinal pathology signs who have scans that lead on to discogram or surgery can “sometimes” actually have very good outcomes when other approaches have not been so successful like trying to resume normal activity or manual/exercise therapy or pain education. I believe it all comes down to the old mistake of treating all non specific low back pain under one umbrella. Lets not disregard MRI all together!

    Neil O'Connell Reply:

    Hi Mark,

    Thanks for your comments. The authors actually do a good job of discussing the possible routes to an iatrogenic effect of MRI scans including those you mention.

    I would defnitely not write off MRI or any other imaging for the low back, but as the clinical guidelines suggest it should be reserved for incidences where the examination throws up indications of specific spinal pathology. I would argue that that alone is “correct use” and anything else is speculative.

    We have no reliable data that MRI scans are useful in the absence of such signs. Many feel the umbrella term of “non-specifc low back pain” to be a mistake as you suggest, but since to date there is no validated and reliable way of better classifying these people, it is the best we have. Many clinicians have their own personal preferred way of classifying them but the validity of any such clinical judgements is suspect at best.

    In terms of better educating patients about the meaning of their scans I see that as reasonable. But it can be very difficult to convince an anxious patient who, like all of us, has an awareness of the curse of “slipped discs” and “degeneration” that these findings are likely to be trivial or unrelated to the problem at hand. You are fighting some powerful and widespread ideas there.

    And as with all things, that some individual patients still do well after MRI scan is neither a validation nor a condemnation of the procedure. To make good decisions we need better data so that we might infer likely outcomes from our clinical decisions. And the results discussed here suggest that there is a risk here that is not offset, on average, by a measurable benefit.

  10. Neil O'Connell says

    Thanks Tasha,

    I agree that the context in which these patients were treated is likely to have had a significant influence on the results here. But even then its a great example of how other influences (insurance policies, financial incentives, clinicians beliefs, patient demands) can all trump what appears to be best practice (with shocking results in this case).

    Now if these results were replicated elsewhere we really would have a big can of worms!

  11. Tasha Stanton says


    Awesome post! It is so funny that you should write on this study. I just went to the Low Back Pain forum in Melbourne last week and saw the presentation of this study and was INTRIGUED! The numbers for early-MRI compared to no-MRI were staggering!

    I’d be hesitant to say that the magnitude of these results would generalize to the regular patients that we see in clinical practice as the results are: 1) from worker’s comp patients and 2) from the US which has a drastically different health care structure. However, I think the message of ‘be careful of who we image’ is spot-on and incredibly important.

    Additionally, one thing that I’m surprised isn’t played out more in the media is how much extra radiation we expose patients to with unnecessary imaging (eg, with unnecessary x-rays and CT scans). Are we causing more problems than we know??

  12. Neil O'Connell says

    Many thanks all, comments much appreciated.

    @lachlan It is hard from the data presented to work out individual paths. What you can (and the authors do) say is that MRI was associated with prolonged disability, hugley increased rates of surgery and increased health costs: more intervention, worse outcomes. How much the outcome is due to the process/ impact of intervention or the unhelpful behavioural changes imposed by the MRI results is hard to say. Given that these results come from a US based work compensation scheme it is fair to say that there may be aspects of that environment that influenced things too. But the effect size blew me away.

    I think its a good argument for resisting patient’s frequent demands for MRI in the same way that GP’s try to resist demands for antibiotics in patients with colds.

  13. Hi Neil,

    Very interesting. Out of curiosity did the patients also have more pain on follow up or is it mainly that they take longer to return to work/ more likely to get surgery due to worrying about a structural problem.

  14. Flavia Di Pietro says

    Great stuff Neil

  15. Really good piece Neil. Thanks