A technical, fully cited reference on diagnosing the source of persistent low back pain: prevalence medians for disc, facet, and sacroiliac joint pain, the likelihood ratios of specific tests, and the population caveats most summaries omit.
Between 85 and 90 percent of low back pain carries the label "non specific," meaning no structural source is confidently identified. The label is honest about diagnostic uncertainty, and clinicians who use it are being more careful than those who confidently blame every backache on a disc. But the label has also become a stopping point, and the research says it should not always be one. In 2023, a systematic review in eClinicalMedicine pooled 62 diagnostic accuracy studies and concluded that for some patients with persistent low back pain, a source diagnosis is possible. This page consolidates that evidence: the prevalence numbers for the three main candidate structures, the specific tests that actually shift the odds, and the limitations that most summaries leave out. It is written as a reference, for clinicians, students, health writers, and patients who want the primary numbers rather than a simplification.
The case for the non specific label is strong and worth stating fairly. Imaging findings correlate poorly with symptoms: in the classic New England Journal of Medicine study, 52 percent of people with no back pain had a disc bulge on MRI, and a 2015 systematic review of 3,110 asymptomatic people found disc degeneration in 37 percent of pain free 20 year olds rising to 96 percent of pain free 80 year olds. If a finding is that common in people who feel fine, its presence in someone who hurts proves little by itself. The 2018 Lancet series took the same position: low back pain is a symptom, serious specific causes are rare, and for most people a precise anatomical diagnosis is not possible.
The 2023 eClinicalMedicine review, by Han, Hancock, Maher, and colleagues, does not overturn that position. It refines it. Pooling 62 studies that compared index tests against reference standards (primarily controlled diagnostic blocks and discography), the authors concluded that informative tests exist for the disc and sacroiliac joint, and one for the facet joint, so "a diagnosis may be possible for some patients." The practical reading: non specific is a description of our tests' limits, not a biological fact about the pain, and the limits are narrower than the label implies.
Across the included studies, the estimated prevalence of each pain source, according to a positive reference standard test, had the following medians:
| Candidate structure | Median prevalence | Interquartile range | Studies |
|---|---|---|---|
| Sacroiliac joint | 53% | 7% | 11 |
| Intervertebral disc (discogenic) | 46% | 26% | 35 |
| Facet (zygapophyseal) joint | 42% | 20% | 14 |
Read this table carefully before quoting it. These percentages come from different studies of different referred populations, each enriched for the structure being investigated, so they cannot be added and do not describe the general population. The reference standards themselves (diagnostic blocks, discography) are imperfect, which is why the review's worst risk of bias domain was the reference standard. And critically: 45 of the 62 studies were conducted in tertiary care, 3 in secondary care, and none in primary care. These are the persistent, referred cases. The right way to quote this table is: "among persistent low back pain patients referred for source workup, each of the three structures is implicated in roughly 40 to 55 percent of the population selected for testing it."
Diagnostic tests are best summarized by likelihood ratios. A positive likelihood ratio (+LR) above 2 meaningfully raises the probability the structure is the source; a negative likelihood ratio (-LR) below 0.5 meaningfully lowers it; values near 1 are noise. Here is what pooling showed:
| Target structure | Test / finding | +LR (95% CI) | -LR (95% CI) |
|---|---|---|---|
| Disc | Modic type 1 changes on MRI | 10.00 (4.20 to 23.82) | 0.84 (uninformative) |
| Disc | Modic type 2 changes on MRI | 8.03 (3.23 to 19.97) | 0.88 (uninformative) |
| Disc | High intensity zone on MRI | 3.10 (2.27 to 4.25) | 0.61 (uninformative) |
| Disc | Centralisation phenomenon on repeated movement testing | 3.06 (1.44 to 6.50) | 0.66 (uninformative) |
| Disc | Annular fissure on MRI | 2.88 (2.02 to 4.10) | 0.24 (informative) |
| Disc | Disc degeneration on MRI | 2.53 (1.57 to 4.07) | 0.15 (informative) |
| Facet joint | Facet uptake on SPECT imaging | 2.80 (1.82 to 4.31) | 0.44 (informative) |
| Sacroiliac joint | Positive provocation test cluster | 2.41 (1.89 to 3.07) | 0.35 (borderline) |
| Sacroiliac joint | Absence of midline low back pain | 2.44 (1.50 to 3.98) | 0.31 (informative) |
| Sacroiliac joint | Radionuclide imaging uptake | 7.33 (1.42 to 37.80) | 0.74 (uninformative) |
Three details deserve emphasis. First, the asymmetry: Modic type 1 changes are strong rule in evidence for discogenic pain (+LR 10) but nearly useless for ruling it out, while generic disc degeneration works the opposite way, weak rule in, decent rule out. A clinician can exploit both directions. Second, the two informative bedside findings, the centralisation phenomenon and the sacroiliac provocation cluster plus pain location pattern, require no imaging at all, which matters given guidelines discourage routine scans. Third, the radionuclide imaging estimate for the sacroiliac joint has a confidence interval running from 1.42 to 37.80, which is a statistical shrug; we include it because the review did, not because you should act on it.
The facet joints illustrate why honest diagnostic writing is hard. The Han review found no informative history, examination, or routine MRI finding for facet joint pain, only SPECT uptake qualified, and SPECT is not a routine test. The literature that does assign confident facet prevalence numbers comes largely from the interventional pain community using controlled diagnostic blocks. The most cited recent figure is from Manchikanti and colleagues (2020): a 34.1 percent prevalence of lumbar facet joint pain in chronic low back pain using controlled blocks, alongside a finding that should be quoted every time the first number is: a single uncontrolled block showed a 67.9 percent prevalence with a 49.8 percent false positive rate. In other words, half of the patients who respond to one diagnostic facet injection are not actually facet cases, which is why single block diagnoses inflate every downstream statistic built on them.
Two caveats on the 34.1 percent itself. It comes from a specialty interventional pain population, not from general practice, and the journal it appears in is closely associated with its authors' own society, a conflict worth knowing about even when the methods are reasonable. The wider controlled block literature puts facet prevalence estimates in a broad range, and 34.1 percent is best treated as one point within it rather than a settled fact.
A reference page should mark its own boundaries, so here are the claims this literature does not support. It does not support ordering an MRI for every backache: the informative findings above were studied in persistent, referred cases, and guidelines from the American College of Physicians continue to recommend against routine imaging in acute low back pain, where most episodes improve within weeks regardless. It does not support the idea that naming a source guarantees a better outcome; the review's authors frame source diagnosis as "potentially guiding targeted treatment," a hypothesis under study, not a proven pathway. And it does not make the asymptomatic imaging problem go away: a Modic type 1 change shifts the odds tenfold within a symptomatic, tested population, which is a different claim from "your MRI finding explains your pain." Pretest context is everything, which is precisely why likelihood ratios, not sensitivities in isolation, are the right currency.
For the persistent low back pain patient, this evidence supports a structured, mostly hands on workup before anyone reaches for advanced testing: pain location mapping (midline versus off midline, since absent midline pain roughly doubles the odds of a sacroiliac source), a provocation test cluster for the sacroiliac joint, repeated movement testing for the centralisation response that points toward discogenic pain, and imaging reserved for cases where the result would change management. It also supports telling patients the truth in both directions: most low back pain never needs a structural label to be treated well, and when pain persists, the label "non specific" is a starting point for a smarter workup, not a verdict. Our patient facing guides to lower back pain, degenerative disc disease, and sciatica translate this material, and our back pain statistics reference covers the epidemiology with the same sourcing standard.
Every figure on this page is sourced to the studies below and was verified against the primary source before publishing; the prevalence medians and likelihood ratios were checked against the full text of the Han 2023 review, not just its abstract. We stated population limits (tertiary care, persistent pain, interventional pain cohorts) and evidence quality problems (imperfect reference standards, single block false positives, journal conflicts) in the same paragraphs as the numbers they qualify. This page is a technical reference reflecting the literature as of the last reviewed date above; it is not medical advice for an individual and does not replace an in person evaluation.
In a 2023 systematic review of 62 diagnostic accuracy studies, the median reference standard prevalence was 53 percent for sacroiliac joint pain, 46 percent for discogenic pain, and 42 percent for facet joint pain, each within the referred populations selected for testing that structure. The figures cannot be added together, and nearly all the underlying studies came from tertiary care.
It means no structural source has been confidently identified, which describes 85 to 90 percent of cases. It is a statement about the limits of our tests, not proof that nothing specific is wrong. Recent evidence suggests a source diagnosis is possible for some patients with persistent pain.
Sometimes, with major caveats. Modic type 1 changes carry a positive likelihood ratio of 10 for discogenic pain in symptomatic tested populations. But disc bulges and degeneration are common in people with no pain at all (52 percent of pain free adults in one classic study had a bulge), so findings must be interpreted against examination, not read as verdicts.
With difficulty. A 2023 systematic review found no informative history, examination, or routine MRI finding for facet joint pain; only SPECT imaging uptake qualified. Diagnostic block studies report around 34 percent prevalence in chronic low back pain with controlled blocks, but a single uncontrolled block carries a 49.8 percent false positive rate.
Two bedside findings are informative: a positive cluster of provocation tests (positive likelihood ratio about 2.4) and the absence of midline low back pain (about 2.4, since SI joint pain typically sits off center, below the beltline). Neither requires a scan, and together they meaningfully shift the odds.
This article is for general education and is not a substitute for an individual evaluation. External links are provided for reference and do not imply endorsement.
