Sixteen Days of Fever: Imaging Further or Treating Now
Sixteen days of unexplained fever, a first-line workup that has found nothing, and a patient who is visibly declining while everyone debates whether the next move is another scan or a treatment.
Walter H., a 58-year-old man, retired from long-haul trucking two years ago and now works three mornings a week at the hardware store because, his wife says, sitting still all day makes him restless. He has spent every autumn weekend for a decade fishing the reservoir with his brother-in-law, and this year he has missed all of them. Sixteen days ago he developed intermittent fevers to 39.1°C, drenching night sweats, and an eight-pound weight loss he did not intend. He has no cough, no dysuria, no rash, no joint pain, and no travel outside the state in over a year. He takes lisinopril for hypertension diagnosed five years ago and is otherwise, by his own account, someone who “hasn't seen a doctor for anything real since the Army.”
The workup so far has been thorough and quiet: three sets of blood cultures drawn on separate days, all negative at five days; CT of the chest, abdomen, and pelvis showing no mass, abscess, or adenopathy; a transthoracic echocardiogram without vegetation, though a transesophageal study was not pursued given the low pretest suspicion. His ESR is 68 mm/hr and CRP 94 mg/L — real, substantial inflammation with nothing yet found to explain it. That gap between a genuinely inflamed patient and a genuinely negative anatomic workup is what FDG-PET/CT is best positioned to close, though the numbers are more modest than the scan's reputation suggests. Bleeker-Rovers et al.'s prospective multicenter cohort, worked up under a structured protocol built around this scan, still ended with no cause found in 51% of patients — though an abnormal FDG-PET was one of the few findings that independently predicted reaching a diagnosis. The figure that applies to Walter is narrower: in Bharucha et al.'s meta-analysis, the yield of FDG-PET/CT over and above a conventional CT — which Walter has already had, and which was negative — was 32%. What it finds disproportionately is large-vessel vasculitis and vertebral infection — neither of which a chest/abdomen/pelvis protocol is built to see.
His baseline health complicates neither reading of the case, which is itself worth stating: no diabetes, no immunosuppressing condition, no prosthetic material anywhere that would raise suspicion for device infection, and his only chronic medication is the lisinopril he has taken without incident for five years. Set against that baseline is a separate, equally real finding from Vanderschueren et al.'s FUO cohort, drawn from patients worked up this thoroughly and this negatively: a meaningful share resolve on their own, undiagnosed, without ever proving dangerous. Walter's wife asked directly today whether “no answer yet” meant something was being missed; the honest answer, borne out by that same cohort, is that it sometimes means the opposite.
Hospital day sixteen, still without a diagnosis
Before anyone starts anything, get the FDG-PET/CT. This is exactly the profile it was built for — a genuinely inflamed patient with a completely negative conventional workup. I want to be honest about the size of the effect: Bharucha's meta-analysis puts its yield beyond a conventional CT at about a third. But a third is not nothing when the alternative is nothing, and what it catches is disproportionately what a chest/abdomen/pelvis CT structurally cannot: large-vessel vasculitis, vertebral osteomyelitis, occult prosthetic-adjacent infection.
If we start empiric antibiotics or steroids first and the scan afterward shows something, we may never know whether the drug already sterilized it — that's not a hypothetical, it's the single most common reason these workups end without an answer.
I'm not arguing against the scan — order it today. I'm arguing against treating “get the PET-CT” as if it resolves the actual bedside problem, which is that he has lost eight pounds and stopped fishing in three weeks, and PET-CT scheduling and read-out in this system routinely runs another four to six days.
You're right that an empiric course could sterilize a focus the scan would otherwise find — but a functionally declining patient sitting through another week of nothing while we wait for a slot isn't a neutral default either; it's its own decision, just an unstated one.
Both of you are treating “do something” and “do nothing” as the only two options, and there's a real third one worth naming plainly: watch, without either escalating imaging urgently or starting a drug, because Vanderschueren's own FUO cohort found a substantial minority resolve entirely on their own, with no diagnosis ever reached and no harm from the wait.
That's not the same claim as “ignore him” — the scan should still happen on its normal urgent-but-not-emergent timeline, and antipyretics and close vitals monitoring cover the actual acute risk. What I'd resist is a bounded empiric antibiotic or steroid trial specifically, because if this turns out to be one of the self-limited cases, we will have exposed him to real drug toxicity for a process that was already ending on its own.
Agreed within the hour: FDG-PET/CT ordered on an urgent (not emergent) basis, acetaminophen for symptomatic control, and no antibiotic or steroid started before the scan result is in hand — the infectious disease physician's sequencing argument carried the room once the hospitalist's concern was addressed by simply expediting the study rather than abandoning the sequence.
Directed therapy follows immediately — biopsy or culture of the identified site before any antibiotic, per the same sterilization logic that held for the empiric trial.
Not agreed: the hospitalist would move to a bounded empiric trial at that point given the now three-week decline; the clinical pharmacologist would still favor continued observation, citing the real base rate of spontaneous resolution — carried forward rather than settled today.