A patient in his early 30s shows up in my office with seven years of back pain. He’s had two lumbar MRIs already. The diagnoses, in order, were muscle strain, disc dehydration, anxiety. He’s seen three primary care physicians and a neurologist. Nobody, in seven years, ordered an X-ray of his sacroiliac joints. When I ask why, the answer is roughly always the same: they didn’t want to expose him to radiation.

This is the part of my practice I find hardest to discuss with colleagues, because the instinct behind it is a decent one. None of us want unnecessary radiation. But somewhere the assumption hardened that “no imaging” equals “safe,” and in suspected axial spondyloarthritis (axSpA) that math doesn’t hold up.

What I think we’re comparing wrong is this. It isn’t radiation versus zero radiation. It’s a small, protocol-dependent dose against years of carrying a treatable inflammatory disease without a diagnosis. RadiologyInfo, the ACR/RSNA patient resource, says so directly: a patient is also at risk when their clinician can’t reach an accurate diagnosis. The second half of that sentence is the part we keep forgetting.

The numbers, in context

A two-view lumbar spine X-ray runs about 1.4 mSv. An AP pelvis is closer to 0.7. Average background radiation in the U.S. is roughly 3 mSv every year, before any medical imaging at all. A full axial series for axSpA comes in somewhere between 1.5 and 4 mSv, depending on protocol, equipment, and patient habitus. Low-dose CT of the sacroiliac joints, in Chahal’s 2018 paper, averaged 0.42 mSv.

Apply ICRP Publication 103’s nominal coefficient of about 5% per Sv to that range, and you get an added lifetime stochastic risk on the order of 1 in 13,000 to 1 in 5,000. Population estimates, not predictions for a particular patient. Small, yes. Zero, no.

The delay is real

Zhao and colleagues (2021) put the pooled mean diagnostic delay in axSpA at 6.7 years. Hay and colleagues (2022) argued that mean delay is inflated by a long right tail and reported medians instead, which came out shorter. Either way, patients are still waiting years between symptom onset and diagnosis.

Some of that wait sits in primary care, where inflammatory back pain looks like mechanical pain at a glance. Some of it sits with us, when we treat imaging as heavier than it is. The result is a long stretch of unexplained symptoms, false reassurance, and a closing window for anti-inflammatory therapy.

AxSpA isn’t a local back problem either. Mathieu’s 2015 meta-analysis put pooled odds ratios around 1.5 to 1.6 for myocardial infarction and stroke in ankylosing spondylitis. Exarchou’s Swedish nationwide cohort showed higher all-cause mortality. Deng’s systematic review reported a modest increase in overall malignancy risk and a stronger signal for multiple myeloma. Whether earlier imaging changes any of those long-term outcomes is plausible but unproven. What we can say without overstating it is narrower: skipping low-dose imaging isn’t, on its own, protective.

MRI isn’t simply a better X-ray

The obvious reply is, “fine, get an MRI instead.” MRI and radiographs answer different questions. MRI is excellent for active inflammation, especially bone marrow edema, and it’s where I start in younger patients, very short symptom duration, pregnancy, or when active inflammation is the central question. X-rays are for structural disease: erosions, sclerosis, ankylosis, syndesmophytes, alignment, transitional anatomy, and the mechanical mimics that quietly get relabeled as inflammatory when no one looks for them.

The major guidelines line up on sequencing, not on ranking the modalities. The ACR Appropriateness Criteria classify sacroiliac (and SI plus area-of-interest spine) radiography as Usually Appropriate for initial imaging in suspected axSpA, with MRI as Usually Appropriate additional imaging. EULAR advises conventional radiography of the SI joints first, with MRI as an alternative first study in selected cases like younger patients or shorter symptom duration. NICE QS170 takes the same approach: X-ray first, MRI if the X-ray doesn’t show sacroiliitis and clinical suspicion remains. NICE adds an explicit caveat that MRI is used first in patients likely to have an immature skeleton.

What I’d ask colleagues to do

I’m not arguing for more imaging. I’m arguing against treating “let’s hold off” as the default.

When the inflammatory back pain features are there (onset under 45, insidious onset, more than three months of pain, morning stiffness, improvement with activity, alternating buttock pain), name what you’re considering. Don’t fold the patient back into mechanical pain because imaging feels heavier than it is.

Pick the modality that answers the question, in the right order. If the question is structural, start with X-ray and add MRI if you still don’t have an answer. If active inflammation is the question, or the patient is very young, or pregnancy considerations apply, start with MRI. If the X-ray is equivocal, the question is still structural, and you have access to a low-dose protocol, low-dose CT.

And read systematically. A vague negative report falsely reassures. A loose positive pushes a patient into the wrong diagnostic bucket. Structural lesions, inflammatory lesions, mechanical mimics, and clinical pattern have to be read together. That is the actual safety intervention. Not “we didn’t image.”

The young man with seven years of back pain finally got the series he should have had years earlier. He had bilateral grade 3 sacroiliitis. Those deferrals cost him seven years.