Evidence-Based Imaging

For Medical Students

What It Shows

The imaging literature answers three different questions: whether a modality detects a lesion, whether detecting it changes management, and whether that change improves the outcome. 5.1.5 covers cost effectiveness.

Most of it answers the first, some answers the second, and little answers the third. 5.21.1 covers the meniscal imaging literature.

That ordering is the single most useful thing to know about this literature, because detection studies are read as if they answered the third question.

Where It Misleads

Detection studies compare a modality against a reference standard, usually arthroscopy, in patients selected for arthroscopy. That selection inflates the apparent performance relative to a clinic population. 5.21.1 covers the meniscal imaging literature.

Management change studies are rarer and are the ones that justify a request. 5.1.3 covers choosing the investigation.

Outcome studies are rarest and are what a cost effectiveness argument actually needs.

No figures appear for any accuracy comparison on this page, under decision 028. The ranking of the modalities is standard teaching and recalled robustly; a percentage is recalled alone and nothing constrains it.

MRCS Level

How Good It Is

In a viva this is the appraisal question and the expected answer separates the three question types.

The expected points are the reference standard, the selection of the population, the distinction between detection and management change, and what the result means for a clinic patient.

Say that arthroscopy as a reference standard carries its own error and is applied to a selected population. 5.21.1 covers the meniscal imaging literature.

Say that a sensitivity from a surgical series does not transfer to a clinic. 5.1.6 covers imaging pitfalls.

Say which question a given paper answers before quoting it.

Where It Misleads

The reasoning expected is that a diagnostic accuracy figure is a property of a test in a population, not of the test alone, so it moves with the population.

A study in a young athletic cohort and one in an older arthritic cohort produce different numbers for the same sequence, and both are correct. 5.6.2 covers meniscal signal grading.

A reference standard that is itself imperfect caps the apparent performance of anything compared against it.

Technology moves faster than the literature, so the sequence studied is often not the sequence being used. 5.18.1 covers quantitative imaging.

FRCS Level

Where It Misleads

At decision level the evidence rarely addresses the decision being made, and pretending otherwise is the error.

The decision is usually whether to operate, and the literature usually reports whether a lesion was seen. Those are two steps apart. 5.1.5 covers cost effectiveness.

Guidelines compress this literature into recommendations, which is useful and hides the question type each recommendation rests on. 5.1.4 covers imaging pathways.

Where the evidence is thin the honest position is to say so and decide anyway, because the patient is in the clinic either way.

The literature on advanced techniques is almost entirely detection and feasibility work, so a clinical claim for them is ahead of its evidence. 5.18.1 covers quantitative imaging.

Fellowship Level

Where It Misleads

At fellowship level this page's own position is the honest illustration of its subject.

This section was written without source access, so it states the comparative standing of the modalities and refuses every figure. That is decision 028 and it is a worked example of the gap between what is known robustly and what is recalled precisely. 5.21.1 covers the meniscal imaging literature.

A reader who wants the numbers has to go to the literature, and the section says where rather than approximating. 5.21.8 covers future directions.

Reference standards are shifting, since arthroscopy is being replaced by imaging in some comparisons, which makes the older accuracy literature hard to compare with the newer. 5.21.1 covers the meniscal imaging literature.

Automated reading is being evaluated against radiologists rather than against outcomes, which repeats the detection and management confusion one level up. 5.18.8 covers artificial intelligence in imaging.

The strongest evidence based statement available about knee imaging is about which question to ask, and that is a reasoning claim rather than a measured one. 5.1.3 covers choosing the investigation.