Clinical library · Clinical reasoning
Imaging in Low Back Pain
The familiar message is that low back pain is over-imaged. That is true. The less familiar finding, from the same review, is that imaging was not performed in 65.6% of patients who presented with red flags. Both failures are real, and they need different fixes.
In one line. Imaging changes management in a small minority of low back pain presentations, is associated with higher costs and healthcare utilisation, and frequently shows findings that are also present in people without pain — but it is also under-used in exactly the patients who most need it.
Physiotherapists in most jurisdictions do not order the imaging. They do influence whether it happens, how the result is explained, and what the patient concludes from it, which is where most of the harm and most of the benefit sit.
Both directions of error
The 33-study review reports overuse and underuse from the same data. Imaging was not performed where appropriately indicated in 65.6% (95% CI 51.8 to 77.2) of patients presenting with red flags, and in 60.8% (95% CI 42.0 to 76.8) of those with clinical suspicion of serious pathology. [1] "Do not image low back pain" is not what the evidence says.
What the evidence shows
| Question | Finding | Source and quality |
|---|---|---|
| How much imaging is inappropriate? | Of patients referred for lumbar imaging: 34.8% (95% CI 27.1 to 43.3) inappropriate by absence of red flags; 31.6% (28.3 to 35.1) by no clinical suspicion of pathology | 33 studies, meta-analysed, GRADE-assessed [1] |
| And of everyone presenting for care? | Imaging inappropriately performed in 27.7% (21.3 to 35.1) judged by duration of episode, 9.0% (7.4 to 11.0) by absence of red flags, 7.0% (1.8 to 23.3) by no clinical suspicion | Same review. The denominator matters enormously — the rate looks very different depending on which population and which criterion [1] |
| How much is under-used? | Imaging was not performed where appropriately indicated in 65.6% (51.8 to 77.2) of patients presenting with red flags | Same review. The authors state that appreciating both underuse and overuse is fundamental to improving practice [1] |
| Does imaging increase costs and use of care? | Moderate-quality evidence that direct costs increase after X-ray (1 RCT, n = 421) and that MRI or imaging is associated with increased healthcare utilisation including injections, surgery and medication (1 RCT and 2 observational studies, n = 3,897) | GRADE-assessed systematic review. Lower-quality evidence pointed the same way for early MRI costs (3 studies, n = 9,535) [2] |
| Does it cause more time off work? | Unclear. Moderate-quality evidence from 2 RCTs (n = 667) showed no significant difference in absence from work; low-quality evidence from 2 observational studies (n = 7,765) showed significantly greater mean absence in MRI groups | The trials and the observational studies disagree, and the trials are the better evidence [2] |
| Are MRI findings meaningless? | No — some are associated with pain. In adults 50 or younger, disc bulge (OR 7.54, 95% CI 1.28 to 44.56), spondylolysis (5.06, 1.65 to 15.53), disc extrusion (4.38, 1.98 to 9.68), Modic 1 changes (4.01, 1.10 to 14.55), disc protrusion (2.65, 1.52 to 4.62) and disc degeneration (2.24, 1.21 to 4.15) were more prevalent in symptomatic individuals | 14 studies, 3,097 individuals (1,193 asymptomatic, 1,904 symptomatic). Note the extremely wide interval on disc bulge [3] |
| Which findings were not associated? | Any Modic change (1.62, 0.48 to 5.41), central canal stenosis (20.58, 0.05 to 798.77), high-intensity zone (2.10, 0.73 to 6.02), annular fissures (1.79, 0.97 to 3.31) and spondylolisthesis (1.59, 0.78 to 3.24) | Same meta-analysis. The stenosis interval spans four orders of magnitude and carries no information [3] |
| Do Modic changes explain pain? | No conclusion can be drawn. Across 15 studies, odds ratios ran from 0.31 (0.1 to 0.95), an inverse association, to 121.4 (11.21 to 1315.08) | Quality poor in 6 studies and moderate in 7; the review states clinicians should not look for the presence or absence of Modic changes to guide management [4] |
| What happens in practice? | Of 8,047 acute low back pain encounters with no red flags, 29% included an imaging order (emergency department 43%, urgent care 18%) and 5% included a physiotherapy order (7% and 4%) | 1,047 clinicians across 49 sites, January to June 2023. Interviews identified the medico-legal assurance imaging provides, complex referral workflows and varied understanding of the physiotherapy discipline as barriers [5] |
Where it misleads
1. "Findings on MRI are just normal ageing" is too strong
The claim is repeated so often that it has become a slogan, and the meta-analysis does not support it wholesale. In adults 50 or younger, six findings were significantly more prevalent in symptomatic than asymptomatic individuals, with odds ratios from 2.24 to 7.54. [3] Five others were not. The correct statement is that some findings carry information and some do not, and that even the associated ones are common enough in pain-free people that a finding does not establish causation in the patient in front of you.
2. A wide confidence interval is not a result
Central canal stenosis came out at an odds ratio of 20.58 with a 95% confidence interval from 0.05 to 798.77. [3] Quoting the point estimate from that would be indefensible. The same applies to the disc bulge interval of 1.28 to 44.56 — statistically significant, but the magnitude is unknown. Read the interval, not the number.
3. The overuse figure depends entirely on the denominator
The same review reports inappropriate imaging at 34.8% and at 7.0%. [1] The first is the proportion of patients already referred for imaging in whom no red flag was present; the second is the proportion of all patients presenting for care in whom imaging was performed without clinical suspicion. Both are correct. Quoting one without the denominator is how a statistic becomes propaganda.
4. Underuse is the finding nobody quotes
65.6% of patients presenting with red flags did not receive imaging that was indicated. [1] A profession that has correctly learned to resist unnecessary imaging can drift into resisting necessary imaging, and the confidence interval here (51.8 to 77.2) is wide but nowhere near zero. See red flags for which findings should trigger that escalation, and note that most individual red flags have weak diagnostic accuracy — which makes clinical judgement more important, not less.
5. Modic changes are the clearest example of a finding outrunning its evidence
Two of the sources on this page disagree. The 2015 meta-analysis found Modic 1 changes more prevalent in symptomatic adults 50 or younger (OR 4.01, 1.10 to 14.55) while any Modic change was not associated (1.62, 0.48 to 5.41). [3] The 2022 review of 15 studies found odds ratios spanning from 0.31 to 121.4 with no consistency by type, level, position or coexisting condition, judged the associations possibly spurious, and concluded clinicians should not use Modic changes to guide management. [4] This page does not resolve that disagreement; it reports both, and the practical implication of the later and more specific review is the more cautious one.
6. Ordering behaviour is driven by things that are not clinical
In 8,047 encounters explicitly free of red flags, 29% still received an imaging order and only 5% a physiotherapy order. [5] The interviews identified medico-legal assurance, complex referral workflows and poor understanding of what physiotherapy does — not disagreement with the guidelines, which clinicians were generally familiar with. Handing a clinician the evidence does not address any of those.
What the evidence supports — and what it does not
Supported
- Not imaging acute non-specific low back pain in the absence of red flags or clinical suspicion. [1][2]
- Expecting higher costs and healthcare utilisation after imaging. [2]
- Imaging when red flags or clinical suspicion are present — where the evidence shows substantial underuse. [1]
- Explaining that many findings occur in people without pain, while acknowledging that some are associated with it. [3]
- Treating Modic changes as not management-guiding. [4]
Not supported
- "All MRI findings are just normal ageing." Six findings were associated with symptoms in under-50s. [3]
- Quoting a single overuse percentage without its denominator and criterion. [1]
- Discouraging imaging in a patient with red flags. [1]
- Claiming imaging causes more time off work. The randomised evidence shows no significant difference. [2]
- Using stenosis, annular fissures, high-intensity zones or spondylolisthesis on imaging as an explanation for pain. [3]
- Expecting education alone to change ordering. Clinicians were already familiar with the guidelines. [5]
How certain is this?
Evidence grade: Moderate.
The appropriateness review covers 33 studies with meta-analysis and GRADE assessment, and reports both directions of error from the same dataset, which is a sign of a review that went looking for what it might not want to find. [1] Its confidence intervals are wide, particularly on the underuse figures.
The costs and utilisation review is GRADE-assessed and explicit about which findings rest on randomised and which on observational evidence — a distinction that changes the conclusion on time off work. [2]
The prevalence meta-analysis is the weakest link in age terms and the most misquoted. It includes 14 studies and 3,097 individuals, its conclusions are restricted to adults 50 or younger, and several of its intervals are too wide to act on. [3] It is cited here because it is the primary source for a claim made constantly in physiotherapy teaching, and the claim is more qualified than the teaching.
The practice data are recent and large (8,047 encounters) but come from one US health system, so the absolute percentages will not transfer to Indian practice; the pattern — imaging ordered six times more often than physiotherapy in red-flag-free acute presentations — probably will. [5]
What would change the grade: appropriateness and underuse data from settings outside high income health systems, and prospective work on what patients actually conclude from being told their scan is normal.
Common questions
Should I tell my patient their MRI findings do not matter?
Not in those words, because it is not accurate and patients can tell. Six findings — disc bulge, spondylolysis, disc extrusion, Modic 1 changes, disc protrusion and disc degeneration — were significantly more prevalent in symptomatic adults 50 or younger. [3] What is defensible is that these findings are also common in people with no pain, that the association does not establish cause in an individual, and that for stenosis, annular fissures, high-intensity zones and spondylolisthesis there was no significant association at all. [3]
Is low back pain over-imaged or under-imaged?
Both, in different patients. Of those referred for lumbar imaging, 34.8% had no red flags; of those presenting with red flags, 65.6% did not get the imaging that was indicated. [1] The review's own conclusion is that recognising both is fundamental to improving practice. A clinic can be guilty of both at once.
Does having a scan make patients worse?
The evidence supports higher costs and more healthcare utilisation — future injections, surgery, medication — on moderate-quality evidence. [2] For time off work it is unclear: two randomised trials (n = 667) found no significant difference, while two observational studies (n = 7,765) found greater mean absence with MRI. [2] The randomised evidence is the stronger of the two.
What about Modic changes?
Do not use them to guide management. A review of 15 studies found odds ratios ranging from 0.31 (an inverse association) to 121.4, with no consistency by type, lumbar level, position or coexisting condition, and judged the findings possibly spurious from publication bias, selective reporting and post hoc analysis. [4] An earlier meta-analysis did find Modic 1 changes specifically more prevalent in symptomatic under-50s. [3] The disagreement is unresolved.
I think my patient needs a scan. What do I do?
Say so to the referring doctor, clearly and with your reasoning. The underuse figures suggest this happens too rarely, not too often. [1] Check red flags and screening for medical referral for what raises the probability of serious pathology and how to escalate it.
Why does the imaging keep getting ordered anyway?
Not because clinicians disagree with the guidelines — they were generally familiar with them. The interview study identified the medico-legal assurance imaging provides, complex workflows for placing physiotherapy referrals, varied understanding of what physiotherapy does, and the absence of feedback loops in emergency and urgent care settings. [5] Three of those four are things a physiotherapy service can actually work on.
References
- Jenkins HJ, Downie AS, Maher CG, et al. Imaging for low back pain: is clinical use consistent with guidelines? A systematic review and meta-analysis. The Spine Journal. 2018 Dec;18(12):2266–2277. doi:10.1016/j.spinee.2018.05.004 PMID 29730460 Systematic review and meta-analysis
- Lemmers GPG, van Lankveld W, Westert GP, et al. Imaging versus no imaging for low back pain: a systematic review, measuring costs, healthcare utilization and absence from work. European Spine Journal. 2019 May;28(5):937–950. doi:10.1007/s00586-019-05918-1 PMID 30796513 Systematic review (GRADE)
- Brinjikji W, Diehn FE, Jarvik JG, et al. MRI Findings of Disc Degeneration are More Prevalent in Adults with Low Back Pain than in Asymptomatic Controls: A Systematic Review and Meta-Analysis. American Journal of Neuroradiology. 2015 Dec;36(12):2394–9. doi:10.3174/ajnr.A4498 PMID 26359154 Systematic review and meta-analysis
- Hopayian K, Raslan E, Soliman S. The association of modic changes and chronic low back pain: A systematic review. Journal of Orthopaedics. 2023 Jan;35:99–106. doi:10.1016/j.jor.2022.11.003 PMID 36438174 Systematic review
- Minick KI, Krueger A, Millward A, et al. Guideline concordant care for acute low back pain: A mixed-methods analysis of determinants of implementation. The American Journal of Emergency Medicine. 2025 Feb;88:162–171. doi:10.1016/j.ajem.2024.11.042 PMID 39637574 Mixed-methods implementation study
About this resource
Using this in clinic
Every figure here is traceable to its source.
This page reports overuse and underuse from the same review, because quoting only the half that supports the profession's preferred message is how a statistic stops being evidence. Where a value could not be verified against the paper it came from, it is not on this page, and the omission is stated rather than filled with a number from a secondary source.