Don't Treat the Scan: When Imaging Lags Behind Infection Recovery

Some of the most uncomfortable moments in infectious diseases come late, when the patient is finally improving and the scan still looks bad.

Picture a patient with Staphylococcus aureus bacteremia and lumbar vertebral osteomyelitis. The fevers have stopped. The blood cultures have cleared. The back pain is still there, but it is clearly better. The C-reactive protein is falling. Then a repeat MRI is obtained at week 5 or 6. The report sounds worse than expected: more endplate destruction, persistent disc space signal, maybe even more extensive inflammatory change. The room changes immediately. Was the treatment inadequate? Is the organism more stubborn than we thought? Do we need a longer course, another biopsy, broader therapy, a surgical consultation, or all of the above?

That reflex is understandable. Imaging feels objective. It gives shape to our anxiety. But it can also pull us off course. In many infections, follow-up imaging does not behave like a microbiologic test of cure. It often reflects slow remodeling, delayed radiographic resolution, residual structural damage, or sterile inflammation. The harder and more useful question is not whether the scan still looks infected. It is whether the patient still behaves as if infection is active.

This problem is not limited to the spine. The same mistake appears when a residual abdominal collection remains visible after source control, when a vegetation persists after endocarditis treatment, or when PET uptake over prosthetic material raises concern without resolving it. The syndromes are different, but the trap is the same: we start treating the picture rather than the patient.

Why bad follow-up imaging pulls us off course

Clinicians do not overreact to scans because they are careless. We overreact because the stakes are real, and because imaging seems more concrete than symptoms, inflammatory markers, or clinical trajectory. A residual cavity on CT looks like something left behind. A worse-looking MRI sounds like progression. A persistent vegetation feels unresolved. It is easier to defend more treatment than less when the image still looks abnormal.

But recovery from infection is not a clean visual process. Dead tissue is resorbed slowly. Bone remodels slowly. Prosthetic surfaces can remain inflamed. Valves can remain damaged. A radiologist may be describing anatomy, edema, or inflammation long after the viable burden of infection has fallen. That does not make imaging unimportant. It means imaging has to be interpreted like any other test: in context, and with some humility about what it can and cannot tell us.

The central lesson from the literature is not that follow-up imaging is useless. It is that routine imaging is usually low-yield in clinically improving patients. Abnormal imaging should change management only when it matches a real clinical concern: persistent fever, recurrent bacteremia, worsening pain, neurologic decline, poor source control, new embolic events, heart failure, drain problems, or another clear sign that infection may still be active.

Intra-abdominal infection: source control matters more than a perfect CT

If there is one syndrome that teaches humility about imaging, it is intra-abdominal infection. Residual collections can remain visible after drainage. Postoperative or postinfectious inflammatory change can linger. Yet the question that usually matters most is not whether the CT looks clean. It is whether source control was achieved.

That distinction matters, because intra-abdominal infection is not the cleanest setting for a broad anti-imaging slogan. A persistent collection can matter a great deal if drainage is incomplete, if the catheter is malfunctioning, if there are undrained loculations, or if a fistula is feeding the problem. The right framework is not "ignore the scan." It is "do not confuse an abnormal scan with treatment failure once source control has truly been achieved."

The best evidence here comes less from radiographic follow-up studies and more from source-control studies. In the STOP-IT trial, Sawyer and colleagues showed that after adequate source control, a fixed short course of antibiotics performed as well as a longer course driven by continued clinical concern. That does not mean every residual cavity is irrelevant. It does mean clinicians should be careful about extending therapy simply because a collection or inflammatory change remains visible on imaging after the patient has otherwise improved.

The procedural literature points in the same direction. In a 2022 study by Sari and colleagues, routine imaging before percutaneous abscess drain removal was often unnecessary in patients who were clinically well, had low drain output, and showed no ongoing signs of infection. Very few required repeat drainage. That is a practical reminder that the drain, the fever curve, the white count, the pain, and the patient's trajectory may tell us more than a routine image obtained for reassurance.

So when should repeat imaging matter in intra-abdominal infection? It matters when the clinical picture stops making sense. A patient who remains febrile after drainage, whose pain is worsening, whose leukocytosis is rising, whose drain output stays high or becomes enteric, or whose source control was partial from the start has given you a new question. In that setting, repeat CT is not defensive medicine. It is the right next test.

The trap is more subtle in the patient who is doing well. We have all seen the scenario: the abscess is smaller but not gone, the report still uses language like "persistent collection," and the temptation is to continue therapy for another week or two just to be safe. Sometimes that is reasonable. Often it is not. If the patient is better, the drain course made sense, the microbiology fits, and the source was controlled, radiographic imperfection alone is a weak reason to keep treating.

Vertebral osteomyelitis is the clearest example of imaging lag

Among common ID syndromes, vertebral osteomyelitis probably offers the strongest evidence that follow-up imaging can look bad long after treatment has begun to work. If there is one place where "don't treat the scan" is most defensible, it is here.

The 2015 IDSA guideline for native vertebral osteomyelitis makes that point directly. In patients with a favorable clinical and laboratory response, routine follow-up MRI is not recommended. The guideline also warns against interpreting worsening bony imaging at 4 to 6 weeks as treatment failure by itself. That is an unusually explicit statement, and it reflects a real mismatch between the way MRI evolves and the way patients recover.

Kowalski and colleagues described this problem well in their AJNR study. Follow-up MRI findings often failed to correlate with clinical status. Bone and disc abnormalities could appear unchanged or even worse, while soft-tissue findings improved and the patient improved with them. Zarrouk and colleagues reported a similar pattern. Even among cured patients, vertebral edema frequently persisted for months, and discal or paravertebral abnormalities could remain visible long after the infection had been brought under control.

That is not surprising if you think about the biology. The spine does not heal quickly. Inflammation, edema, granulation tissue, and structural damage may persist on MRI well beyond the point when viable infection has meaningfully declined. In fact, bony destruction may become more radiographically conspicuous as the disease reveals damage that was already underway before treatment started. What the MRI often captures is not live infection in real time, but the aftermath of infection in a tissue that remodels slowly.

This is where many of us get into trouble. We order repeat imaging because we want certainty, then receive a report that documents persistent abnormality, and finally feel obligated to respond to it. The patient may be clinically improved, the inflammatory markers may be falling, and the bacteremia may be gone, but the image becomes the loudest voice in the room.

That does not mean follow-up MRI has no role. It matters a great deal when the patient has new or worsening neurologic findings, persistent or worsening pain that does not fit the expected course, recurrent bacteremia, rising inflammatory markers after an appropriate interval, or concern for enlarging epidural or paraspinal soft-tissue disease. It also matters when the original source control question was never settled. But those are clinical triggers for imaging. They are not arguments for routine imaging in a patient who is recovering.

One practical nuance is worth emphasizing. In vertebral osteomyelitis, soft-tissue evolution is often more informative than persistent marrow or endplate changes alone. An improving epidural component, a shrinking paraspinal collection, and a better clinical course often matter more than residual bony destruction on the report. That distinction helps us read the MRI as a biologic story rather than as a simple positive-or-negative test.

Endocarditis creates a different imaging problem, but the same cognitive trap

Endocarditis is different from vertebral osteomyelitis in one important way: we often image not just to look for persistent infection, but also to define complications. Echocardiography is asked to tell us about vegetation size, valve dysfunction, abscess formation, prosthetic failure, hemodynamics, and sometimes embolic risk. That makes the follow-up question more complicated. Still, the same mistake shows up here too. We see persistent abnormalities and assume the infection itself must still be active.

Older echocardiographic literature already warned against that assumption. In a classic study by Vuille and colleagues, vegetations often persisted after otherwise successful medical therapy. Many did not shrink dramatically by the end of treatment. Persistence alone did not automatically predict bad late outcomes. In other words, a remaining vegetation is not the same thing as uncontrolled infection.

More recent data have raised the same question about routine end-of-therapy echocardiography. In the 2020 cohort reported by Virk and colleagues, patients who had no new or worsening signs or symptoms near the end of treatment rarely had follow-up echo findings that changed management. That does not make end-of-therapy echocardiography wrong. The AHA statement still supports it largely as a new baseline in many cases. But it does challenge the idea that every residual abnormality discovered at the end of therapy is clinically actionable.

This distinction matters at the bedside. A persistent vegetation may represent residual infected material, but it may also represent sterilized thrombotic debris, damaged valve architecture, or the structural scar left by infection. The echo may be showing anatomy after endocarditis, not active endocarditis itself. The same is true of residual regurgitation or other abnormalities that reflect what infection did, not what infection is still doing.

What should push us back toward repeat imaging in endocarditis? Recurrent fever or bacteremia should. New heart failure should. A new murmur, conduction abnormality, embolic event, concern for abscess, prosthetic dysfunction, or clinical relapse should. Those are all situations where imaging answers a live diagnostic question. But in the patient who has completed therapy and is clinically stable, persistence by itself is a thin reason to restart treatment or assume failure.

PET/CT is useful, promising, and very easy to overinterpret

PET/CT deserves its own section because it has changed the diagnostic conversation around prosthetic valve and device-related endocarditis, but it has not solved the follow-up problem cleanly.

In diagnosis, PET/CT can be extremely helpful, especially when prosthetic material blurs the limits of echocardiography. The ESC endocarditis guideline incorporated that reality years ago. The problem is that a strong diagnostic test is not automatically a strong test-of-cure tool. Inflammation over prosthetic material can persist. Postoperative change can persist. Metabolic activity is not synonymous with viable infection.

That ambiguity shows up in the follow-up literature. Bucy and colleagues reported that in medically treated prosthetic valve endocarditis, heterogeneous prosthetic uptake often persisted on follow-up PET/CT. That is a useful observation, not because it tells us persistent uptake is benign, but because it reminds us that uptake alone is not self-interpreting. A hot prosthetic valve on PET months into treatment is not the same thing as a positive blood culture.

This is where PET/CT can create a false sense of precision. The image looks advanced. The signal looks biologic. The report sounds quantitative. But if the patient's syndrome is not pointing in the same direction, the scan may clarify less than we hope. PET/CT is often most useful when there is a real unresolved question: relapse versus sterile postoperative change, persistent occult infection around prosthetic material, or ongoing concern that cannot be settled by clinical course and echocardiography alone.

That is a narrower role than many of us want. But it is probably the honest one. PET/CT should be interpreted as part of a clinical argument, not as a stand-alone verdict.

When should repeat imaging actually change management?

The practical value of all this literature is not that it gives us permission to ignore abnormal scans. It gives us a framework for deciding when a repeat scan is worth acting on.

Across these syndromes, repeat imaging becomes high-value when the patient gives you a new question:

• persistent or recurrent fever
• recurrent bacteremia or microbiologic relapse
• worsening focal pain or a course that has stopped improving
• new neurologic findings in vertebral osteomyelitis
• rising inflammatory markers that fit clinical nonresponse
• inadequate or uncertain source control in intra-abdominal infection
• drain dysfunction, persistent high output, or concern for fistula
• new heart failure, embolic events, prosthetic dysfunction, or conduction changes in endocarditis

What does not count as a strong stand-alone trigger? A scan that still looks abnormal in a patient whose clinical trajectory says the opposite.

That last sentence sounds simple, but it is often the hardest part. We are trained to distrust ambiguity, and residual imaging abnormality is a very visual form of ambiguity. It is tempting to act on the scan because it feels safer than tolerating uncertainty. But more antibiotics are not always more safety. Sometimes they are just a way to quiet our discomfort.

The deeper skill is learning what treatment failure really looks like

In all three syndromes, the real challenge is definitional. What are we calling failure?

If failure means persistent active infection, then it should usually declare itself through more than radiographic persistence alone. It should show itself through the return of bacteremia, the persistence of fever, the failure of pain or function to improve, the inability to control the source, or the emergence of a new complication. A lingering abnormal image may accompany those things. It does not reliably substitute for them.

That is why the phrase "don't treat the scan" is useful only if we hear it correctly. It does not mean imaging is irrelevant. It means imaging is subordinate to syndrome, timing, microbiology, source control, and the patient's actual arc of recovery. Sometimes the image is the key to the case. Sometimes it is just a record of the damage that infection has already done.

I think that is why follow-up imaging is so unsettling in ID. It forces us to distinguish active disease from residual injury, and those are not the same thing. A worse-looking MRI may still be a recovery MRI. A residual abdominal collection may still be a controlled infection. A persistent vegetation or PET signal may still represent aftermath rather than failure. The task is not to be cavalier about those possibilities. It is to weigh them honestly.

In practice, that means asking a stricter question before responding to a scan: what exactly is this image explaining that the rest of the case is not? If the answer is clear, the scan is helping. If the answer is only that the picture remains abnormal, we should probably slow down before changing treatment.

Practical Takeaways

• Follow-up imaging is a problem-solving tool, not a routine proof-of-cure test.
• In intra-abdominal infection, the key question is usually whether source control was achieved, not whether every residual cavity has disappeared.
• Vertebral osteomyelitis offers the strongest evidence that MRI can remain abnormal or even look worse despite clinical improvement.
• In endocarditis, persistent vegetations and some structural abnormalities can remain after successful therapy and do not automatically mean active infection.
• PET/CT can help in selected prosthetic or device-related cases, but persistent uptake is not self-interpreting and should not be treated as proof of failure.
• Repeat imaging is highest-yield when the patient gives you a new reason to ask a new question.

Musical Coda

A single image can trap our thinking long after the patient has started to move forward.

References

1. Berbari EF, Kanj SS, Kowalski TJ, et al. 2015 Infectious Diseases Society of America (IDSA) Clinical Practice Guidelines for the Diagnosis and Treatment of Native Vertebral Osteomyelitis in Adults. Clin Infect Dis. 2015;61(6):e26-e46. doi:10.1093/cid/civ482.
2. Kowalski TJ, Berbari EF, Huddleston PM III, Steckelberg JM, Osmon DR. Do follow-up imaging examinations provide useful prognostic information in patients with spine infection? AJNR Am J Neuroradiol. 2007;28(4):693-699.
3. Zarrouk V, Feydy A, Salles F, et al. Imaging does not predict the clinical outcome of bacterial vertebral osteomyelitis. Rheumatology (Oxford). 2007;46(2):292-295. doi:10.1093/rheumatology/kel221.
4. Sawyer RG, Claridge JA, Nathens AB, et al. Trial of short-course antimicrobial therapy for intraabdominal infection. N Engl J Med. 2015;372(21):1996-2005. doi:10.1056/NEJMoa1411162.
5. Sari MA, Heiss KF, Bairdain S, et al. Is routine imaging necessary prior to percutaneous abscess catheter removal? Abdom Radiol (NY). 2022.
6. Solomkin JS, Mazuski JE, et al. IDSA Clinical Practice Guideline Update on Complicated Intra-abdominal Infections. 2024.
7. Baddour LM, Wilson WR, Bayer AS, et al. Infective Endocarditis in Adults: Diagnosis, Antimicrobial Therapy, and Management of Complications. Circulation. 2015;132(15):1435-1486. doi:10.1161/CIR.0000000000000296.
8. Habib G, Lancellotti P, Antunes MJ, et al. 2015 ESC Guidelines for the management of infective endocarditis. Eur Heart J. 2015;36(44):3075-3128. doi:10.1093/eurheartj/ehv319.
9. Vuille C, Nidorf M, Weyman AE, Picard MH. Natural history of vegetations during successful medical treatment of endocarditis. Am Heart J. 1994;128(6 Pt 1):1200-1209. doi:10.1016/0002-8703(94)90729-7.
10. Virk A, Sohail MR, Stulak JM, et al. End-of-therapy echocardiography may not be needed in all patients with endocarditis. Open Forum Infect Dis. 2020;7(3):ofaa021. doi:10.1093/ofid/ofaa021.
11. Bucy L, et al. Real world experience of therapeutic monitoring of medically treated prosthetic valve infective endocarditis by 18F-FDG-PET/CT. J Nucl Cardiol. 2023.