Ten MRI centers, one spine, nothing in common

A second opinion is only as good as what you hand it. If what you hand it is a report, you are handing it one reader’s interpretation — and the research says that interpretation depends heavily on which building you walked into.

As a neurosurgeon, I read spine imaging myself, and I read the reports that come with it. Those are not the same thing, and the gap between them is the subject of this post — because it is the single best reason I know to carry your actual images, not just the paperwork, into any second opinion.

One woman, ten scanners, three weeks

In 2017 a group of researchers in The Spine Journal did something that sounds almost too simple. They took one 63-year-old woman with low back pain and symptoms running down her right leg, and over three weeks they sent her for a lumbar spine MRI at ten different MRI centers. Same patient. Same spine. Nothing about her anatomy changed in three weeks. Then they compared the ten reports.

It is one patient, and the reference standard was the study authors’ own consensus reading, so I wouldn’t hang a precise error rate for all of radiology on it. But you don’t need a precise rate for the point to land. The authors’ own conclusion was that both the imaging center and the radiologist reading the scan can affect the diagnosis — and therefore the treatment.

Same woman. Same spine. Three weeks. Ten reports, and not a single finding that all ten agreed on.

This is not a story about bad radiologists

I want to be careful here, because the lazy version of this post — radiologists get it wrong — is both unfair and wrong. Much of this evidence was produced by radiologists, studying their own field, in their own journals. That is what a serious specialty does.

Reading a spine MRI is hard. It is hundreds of images, interpreted by a person, often quickly, often without the patient’s history in front of them. A review in Insights into Imaging estimates the ordinary day-to-day discrepancy rate in radiology at 3–5% of studies reported, and much higher in targeted studies that go looking. Scanners differ. Image quality differs. Readers differ in where they set the bar for mentioning something.

So the honest summary is not that radiology over-calls, or that it under-calls. It is that it varies, in both directions — and variation is harder to live with than a consistent bias, because you can’t correct for it. In the ten-center study the bigger problem was misses. But there is a second problem that runs the other way, and it matters just as much to patients.

A finding is not a cause

Even a perfectly accurate report can mislead you, because the things it lists are extremely common in people who feel fine.

The authors’ conclusion was that many of these imaging features are likely part of normal aging and not associated with pain. Read that next to your own MRI report. A report full of findings is not a report full of causes. “Disc bulge at L4–5” in a 60-year-old may be the reason for their leg pain, or it may be something most of their neighbors have too. Deciding which is the clinical work — and it is done by putting the images next to the history and the examination, not by counting lines on a PDF.

A second opinion is only as good as what you hand it

Second opinions change things. How often depends entirely on who is asking and where, so here are the numbers with their fine print attached.

That 88% is not a rate at which diagnoses are wrong. These were patients already selected for referral to a major academic center — the hard, unresolved cases — and “refined” often means a working diagnosis was made more specific, not that it was mistaken. It tells you that a careful second look at a selected, difficult group changes the picture often. It does not tell you that your doctor is probably wrong.

Imaging rereads point the same way. At a cancer center that re-read outside abdominal CT and MRI studies, 34% of the second interpretations were discrepant with the original report, and of the 88 confirmed discrepancies, 48% led to a change in treatment. And in a São Paulo program where an insurer sent patients already recommended for spine surgery for a second assessment, only 15.5% of 425 patients got the same surgical recommendation back. That program was arranged by the insurer, which had its own interest in the answer, and disagreement on its own doesn’t tell you which opinion was right — but it does tell you the first answer is rarely the only defensible one.

Now put all of this together. If reports vary this much between readers, then a second opinion built on the first report is not really a second opinion. It is a second reader of the first reader’s summary. The thing that lets a new doctor actually disagree — or confirm, with confidence — is the study itself.

If the report is one reader’s interpretation, the images are the evidence. Bring the evidence.

What I’d actually do

1. Ask for the images, not just the report. The imaging center can give you the study itself (the DICOM files), and under HIPAA you have a right to a copy. A PDF report is not the scan.
2. Bring your older studies too. The most useful thing a prior scan does is answer “was this already there?” A finding that hasn’t changed in five years reads very differently from one that is new.
3. Read the findings list as a description, not a diagnosis. Many items on a spine report are normal for your age. Which one, if any, explains your symptoms is a conversation with your physician.
4. Ask the second-opinion doctor whether they looked at the images themselves. It is a fair question, and most will tell you plainly.
5. Remember the variation runs both ways. A different reading is not automatically the right one. The goal is a decision made from the evidence, not a search for the report you prefer.

This article is general information, not medical advice. UploMD does not diagnose, interpret imaging, or recommend treatment. Nothing here should be used to make a medical decision on your own — if something in your imaging or your report concerns you, discuss it with a physician.