Fevers, falling counts and a negative infectious workup after immunotherapy — an interactive case that turns on a single timing detail.
August 30, 2026 · 18 min
1. The Presentation
A 60-year-old woman with stage II triple-negative breast cancer presents with a week of fevers, rigors, malaise, a dry cough and nausea. She has intermittent lightheadedness, headaches and neck pain when febrile. Two months ago she completed neoadjuvant carboplatin, paclitaxel and doxorubicin; one month ago she underwent lumpectomy and sentinel-node biopsy. She is also enrolled in a trial of pembrolizumab, last dosed six days ago.
What is the highest-yield next step?
✗ Not quite — the correct answer is highlighted above
Empirical cover is entirely reasonable in a febrile patient with recent chemotherapy — but starting antibiotics is not the same as thinking. Cultures alone will not explain this illness, and the history is the cheapest thing you will do all admission.
✓ Correct
She is immunocompromised and recently instrumented, so the differential spans infection, surgical complication, malignancy and drug toxicity. A structured exposure history — animals, food, travel, ticks, contacts — is the fastest way to generate specific, testable hypotheses before anchoring on the surgical site.
✗ Not quite — the correct answer is highlighted above
Her counts had already normalised after the last cycle, and chemotherapy-associated neutropenic fever would be expected sooner after treatment. Attribution without a workup is unsafe here.
✗ Not quite — the correct answer is highlighted above
Worth imaging eventually, but her postoperative axillary tenderness had already resolved before she presented. Committing to this anchor first risks missing a systemic process.
2. The Exposure History
She takes no regular medications and has no allergies. She lives in a suburban part of the north-eastern United States with no animal exposures. She hikes and bikes outdoors but reports no insect bites. Her son, who lives with her, recently travelled to the United Arab Emirates without becoming unwell. She has eaten no raw, unpasteurised or imported food, but recently handled raw goat and lamb meat from a local butcher.
Which hypotheses does this history generate?
✗ Not quite — the correct answer is highlighted above
It is not. Handling raw goat and lamb raises brucellosis and toxoplasmosis, and she lives in a region endemic for tickborne infection. Both are specific and both are testable.
✓ Correct
These are exactly the falsifiable hypotheses an exposure history should produce. Worth noting in advance — all of them will ultimately test negative. The value of the exposure history is not that it always yields the answer, but that it lets you close doors deliberately rather than by omission.
✗ Not quite — the correct answer is highlighted above
He remained well. Secondary transmission without an index illness is a weak thread — worth noting, low yield.
✗ Not quite — the correct answer is highlighted above
Borrelia, babesia, anaplasma and ehrlichia are endemic and belong on the list — she is later treated empirically for them — but she reports no bites, and this is one hypothesis among several rather than the front-runner.
3. The Examination
She appears comfortable. Temperature 38.9°C, heart rate 120, blood pressure 100/63 mmHg, oxygen saturation 99% on air. There is no cervical, submandibular or supraclavicular lymphadenopathy, no hepatosplenomegaly, and the chest is clear. The lumpectomy incision is healing well without surrounding erythema, fluctuance or tenderness. On skin examination there is a small (<1 cm) erythematous, tender but non-fluctuant papule on the scalp, present four days, which she had not noticed.
How should the scalp papule be interpreted?
✗ Not quite — the correct answer is highlighted above
Dismissing a new skin lesion in an immunocompromised host is precisely how a clue gets lost. That the patient had not noticed it makes it easy to overlook, not unimportant.
✗ Not quite — the correct answer is highlighted above
The absence of fluctuance or purulent drainage argues against a pyogenic collection. This is the one conclusion the lesion does not support.
✓ Correct
A solitary papule here could be atypical erythema migrans, a cutaneous metastasis, Sweet syndrome (usually multiple lesions, so less likely with one) — or VZV reactivation, which deserves particular thought given her immunosuppression, even without pain or a dermatomal rash. Keep it on the list. It returns at the end of this case.
✗ Not quite — the correct answer is highlighted above
Cutaneous deposits belong in the differential, but would be unusual as a solitary tender papule in stage II disease — and would not explain a week of fevers.
4. The Blood Counts
White-cell count 1860/µL, haemoglobin 9.2 g/dL, platelets 66,000/µL. One month earlier these were 4670/µL, 11.2 g/dL and 162,000/µL. The neutrophil count is 1390/µL (normal 1920–7600), lymphocytes 350/µL (720–4100) and monocytes 100/µL (160–1100). Metabolic panel, prothrombin time, partial thromboplastin time, fibrinogen and lactate are all unremarkable.
Which feature is most informative?
✗ Not quite — the correct answer is highlighted above
At 1390/µL this is mild. Neutropenia is generally defined as an absolute count below 1500, and the 1000–1500 band does not substantially raise infection risk. It cannot carry the weight of this illness.
✓ Correct
All three lineages have fallen sharply from documented normal values four weeks earlier, and her counts had already recovered after the last chemotherapy cycle. New pancytopenia in a febrile patient points to marrow suppression, peripheral consumption, or a systemic process — not to residual chemotherapy effect.
✗ Not quite — the correct answer is highlighted above
A haemoglobin of 9.2 in isolation would be unremarkable after chemotherapy. What matters is that it is falling in parallel with the other two lineages.
✗ Not quite — the correct answer is highlighted above
Normal PT, PTT and fibrinogen argue against disseminated intravascular coagulation, but many consumptive and hyperinflammatory states leave these intact early. This does not close the door — and the fibrinogen becomes interesting again later.
5. The Infectious Workup
C-reactive protein is 248.9 mg/L (normal <3.0) and the ESR is 34 mm/h. Nasopharyngeal PCR for Covid-19, influenza, parainfluenza, adenovirus and respiratory syncytial virus is negative, as are sputum, throat and blood cultures. Serology for pertussis, mycoplasma, brucella and toxoplasma is negative; serum PCR for Epstein–Barr virus, parvovirus, cytomegalovirus and toxoplasma is negative. Beta-D-glucan and galactomannan are negative. Urine culture grows only insignificant commensals.
How should this comprehensively negative workup change your thinking?
✗ Not quite — the correct answer is highlighted above
Too strong. Culture-negative infection is common, and withdrawing cover from a febrile patient with progressive cytopenias is unsafe. A negative test lowers a probability; it rarely abolishes it.
✓ Correct
The right move is Bayesian rather than binary. Each negative result lowers the prior for infection, while a CRP of 249 confirms intense inflammation without naming its source. Keep covering empirically, but start actively working the non-infectious differential.
✗ Not quite — the correct answer is highlighted above
CRP rises and falls considerably faster than ESR, so discordance between them is common and non-specific. It quantifies inflammation; it neither localises nor identifies it.
✗ Not quite — the correct answer is highlighted above
The differential has already been searched broadly and returned nothing. Repeating the same category of test has diminishing returns — the next useful steps are imaging and a change of frame.
6. Cross-Sectional Imaging
CT of the chest, abdomen and pelvis with contrast shows a persistent fluid collection of roughly 4 × 3 cm in the left axilla, at the site of her lumpectomy and node biopsy. It is slightly larger than on previous imaging but shows no wall thickening, no enhancement, no internal gas and no surrounding fat stranding. The spleen measures 13.7 cm, having been under 11.5 cm one month earlier. There are no other findings.
Which finding matters most?
✗ Not quite — the correct answer is highlighted above
It is larger, but every radiological feature that would suggest infection is absent — no wall thickening, no enhancement, no gas, no fat stranding — and she has no local pain or tenderness. This is a seroma, an expected consequence of axillary surgery. Anchoring here is the trap this case sets.
✓ Correct
The spleen has grown from under 11.5 cm to 13.7 cm in a single month. Splenomegaly in isolation is non-specific — infective, neoplastic and inflammatory causes all produce it — but a new interval change in a febrile patient with pancytopenia is a genuine systemic signal, and it is the finding that redirects the case.
✗ Not quite — the correct answer is highlighted above
It lowers the probability without excluding it — lymphoma can be extranodal or marrow-based. Flow cytometry and molecular studies address this question far more directly than a CT report.
✗ Not quite — the correct answer is highlighted above
A seroma after axillary surgery is indeed expected. An interval increase in spleen size is not a consequence of breast surgery.
7. Empirical Therapy, and Deterioration
Vancomycin and cefepime are started on admission. She remains febrile. On day 3 cefepime is changed to piperacillin–tazobactam for improved anaerobic cover, and doxycycline is added to treat empirically for tickborne infection. Despite broad-spectrum cover the fevers continue, and on hospital day 4 she becomes symptomatically hypotensive.
What does failure to respond to broad-spectrum antibiotics, now with hypotension, most suggest?
✗ Not quite — the correct answer is highlighted above
Worth considering, but beta-D-glucan and galactomannan were both negative and there is no other evidence of invasive fungal disease. Escalating spectrum indefinitely against a negative workup is a well-worn way to lose several days.
✗ Not quite — the correct answer is highlighted above
The only candidate collection has no imaging features of infection whatsoever. Draining a seroma to chase a fever is low-yield and delays the real diagnosis.
✓ Correct
Fevers unresponsive to broad cover, an exhaustively negative microbiological search, progressive cytopenias, new splenomegaly and now hypotension together fit a systemic inflammatory syndrome rather than an occult abscess. The hypotension is the cue to reframe the problem, not to escalate antibiotics again.
✗ Not quite — the correct answer is highlighted above
Possible in principle, but drug fever does not usually produce progressive pancytopenia, new splenomegaly or hypotension — and she was febrile for a week before any antibiotic was given.
8. Narrowing the Differential
There is no evidence of active infection, no new medication to blame, and no sign of cancer progression on imaging. Attention turns to inflammatory syndromes recognised to follow immune checkpoint inhibition. She has received multiple cycles of pembrolizumab across roughly six months, the most recent six days before presentation.
Which investigations would you send now?
✗ Not quite — the correct answer is highlighted above
Already done, already negative. Repetition at this point is reflex rather than reasoning.
✓ Correct
These four are the laboratory limbs of the diagnostic criteria for the hyperinflammatory syndrome now at the top of the list. Ferritin in particular is cheap, fast, and can be spectacularly abnormal — it is often the test that converts a suspicion into a diagnosis.
✗ Not quite — the correct answer is highlighted above
It may well be needed and can demonstrate haemophagocytosis, but the serum panel is faster, less invasive, and can establish the diagnosis without it. Marrow findings are one criterion among several, not a prerequisite.
✗ Not quite — the correct answer is highlighted above
Reasonable later — occult malignancy, particularly lymphoma, is the commonest driver of such syndromes in adults, and flow cytometry is already pending. But it will not explain an acutely deteriorating patient today.
9. The Confirmatory Panel
Her hypotension resolves with fluid boluses, but the white-cell count and haemoglobin continue to fall, to lows of 1010/µL and 7.0 g/dL. Ferritin is 37,956 µg/L (normal 13–150). Triglycerides are 296 mg/dL (normal 35–150). Fibrinogen is 222 mg/dL (normal 200–400). Soluble CD25 is 4268 pg/mL (normal 175–858), and interleukin-6 is 56.2 pg/mL (normal <7.1). Flow cytometry and molecular testing show no haematological cancer.
Which statement about this panel is correct?
✗ Not quite — the correct answer is highlighted above
A common misreading. The relevant criterion is hypertriglyceridaemia or hypofibrinogenaemia — either satisfies it. Her triglycerides are raised, so the criterion is met despite a fibrinogen within the normal range.
✓ Correct
This is more than 250 times the upper limit of normal. Extreme hyperferritinaemia is not perfectly specific — adult-onset Still’s disease, severe sepsis and iron overload can all raise it — but at this magnitude, in this clinical context, it is powerfully suggestive, and it is itself one of the diagnostic criteria.
✗ Not quite — the correct answer is highlighted above
It is clearly elevated, but the established criteria were validated using a different functional assay, and there is no accepted conversion between the two. It supports the diagnosis; whether it formally satisfies that particular criterion is genuinely uncertain.
✗ Not quite — the correct answer is highlighted above
IL-6 rises in both conditions and cannot discriminate between them. The feature that separates them is timing — which the next two steps address.
10. Applying the Criteria
The HLH-2004 criteria permit a diagnosis when a patient carries a recognised genetic mutation, or fulfils five of eight features: fever; splenomegaly; cytopenias affecting at least two lineages; hypertriglyceridaemia or hypofibrinogenaemia; haemophagocytosis in marrow, spleen or lymph node; low or absent natural-killer-cell activity; hyperferritinaemia; and raised soluble CD25.
Applying these criteria to her, how many are clearly met?
✗ Not quite — the correct answer is highlighted above
An undercount. Fever, splenomegaly, cytopenias across three lineages, raised triglycerides and extreme hyperferritinaemia already make five — before invoking the contested soluble CD25.
✓ Correct
Fever, splenomegaly, cytopenias affecting all three lineages, hypertriglyceridaemia and hyperferritinaemia are unambiguous — five, which meets the threshold. Soluble CD25 is raised but cannot be mapped cleanly onto the criteria, and marrow haemophagocytosis and NK-cell activity were never assessed. The diagnosis stands without them.
✗ Not quite — the correct answer is highlighted above
Haemophagocytosis is one of eight criteria, not a requirement, and it is frequently absent early in the illness. Insisting on marrow confirmation delays treatment in a patient who already qualifies.
✗ Not quite — the correct answer is highlighted above
A recent study found that removing NK-cell activity from the criteria did not reduce diagnostic accuracy, raising real questions about how much it contributes. It is not a prerequisite.
11. Which Trigger?
The two diagnoses that best fit an inflammatory syndrome after pembrolizumab are HLH and cytokine release syndrome. Both are recognised complications of checkpoint inhibition, and both can produce fever, cytopenias and hypotension with raised inflammatory cytokines.
Which feature best favours HLH over cytokine release syndrome here?
✗ Not quite — the correct answer is highlighted above
Extreme hyperferritinaemia does favour HLH, but it is not the cleanest discriminator available in this case, and it is not wholly specific.
✓ Correct
Cytokine release syndrome characteristically begins within days of starting immunotherapy. She had received multiple cycles across roughly six months. Checkpoint-inhibitor-associated HLH appears considerably later — in one series of 18 patients the median onset was 6.7 weeks after treatment began, with a range of 3 to 16 weeks. Timing is the discriminator.
✗ Not quite — the correct answer is highlighted above
Distributive shock occurs in both syndromes. It marks severity, not aetiology.
✗ Not quite — the correct answer is highlighted above
IL-6 is elevated in both — indeed it is central to cytokine release syndrome — so it cannot separate them.
12. Management
A diagnosis of checkpoint-inhibitor-associated HLH is made. Evidence for managing this entity is thin — there are no randomised trials, and practice is extrapolated from case series and from other forms of secondary HLH.
What is the appropriate initial management?
✗ Not quite — the correct answer is highlighted above
Continuing the trigger is the error. Anakinra has a genuine role in this syndrome, but the offending agent must stop first.
✓ Correct
Withdraw the trigger, then suppress the inflammatory response. She received intravenous methylprednisolone at 1 mg/kg daily, with immediate resolution of fever, easing of the cytopenias and a fall in ferritin. The speed of that response is itself characteristic of the syndrome.
✗ Not quite — the correct answer is highlighted above
Etoposide is reserved for refractory disease in this setting. Reaching for it first exposes the patient to unnecessary cytotoxicity when glucocorticoids alone often suffice.
✗ Not quite — the correct answer is highlighted above
It is not self-limiting. Untreated, the inflammatory response can progress to distributive shock and multiorgan failure — and she was already hypotensive on day 4.
13. The Scalp Lesions, Revisited
She improves rapidly, transitions to oral prednisone on day 9, and is discharged on day 11 with a slow taper. After discharge, further scalp lesions appear. These are biopsied and are consistent with shingles, and she is treated with valaciclovir.
Was varicella-zoster reactivation the trigger for her HLH?
✗ Not quite — the correct answer is highlighted above
Superficially attractive, and the timing of that first papule is genuinely suggestive. But two features of the subsequent course argue against it — see the correct answer.
✓ Correct
Her fever and cytopenias resolved with corticosteroids before valaciclovir was given, and new lesions emerged during immunosuppression — a pattern that fits reactivation caused by treatment rather than a preceding trigger. Where varicella does drive HLH, it usually follows primary infection or vaccination rather than reactivation. Pembrolizumab remains the likely cause.
✗ Not quite — the correct answer is highlighted above
That would have been useful, and its absence is a genuine gap. But the clinical sequence — steroid response preceding antivirals, then new lesions under immunosuppression — is already strong enough to support a judgement.
✗ Not quite — the correct answer is highlighted above
Even had VZV contributed, resuming the agent most likely responsible would be unwise. Pembrolizumab was discontinued.
14. Outcome and Take-Aways
She continued to do well, with prednisone tapered by 10 mg weekly, and pembrolizumab was discontinued permanently.
HLH is inappropriate immune activation leading to persistent macrophage activity, and in severe cases to distributive shock and multiorgan failure. The mechanism explains the criteria almost line by line: activated T lymphocytes and macrophages release proinflammatory cytokines, producing fever; the resulting inflammation suppresses the marrow, producing cytopenias; TNF-α inhibits lipoprotein lipase, which normally clears triglycerides, producing hypertriglyceridaemia; and macrophages secrete enzymes promoting fibrin degradation while synthesising excess ferritin. Elevated CXCL9 (a marker of interferon-γ production) and IL-18 are also associated.
Primary HLH presents mainly in childhood and reflects mutations in genes governing T-cell or NK-cell cytotoxicity or inflammasome activity — PRF1, UNC13D, STXBP2, STX11. Secondary HLH accompanies cancer, infection or autoimmune disease. Among adults admitted non-electively with HLH, malignancy is the commonest associated condition, with lymphomas predominating; infections come next, most often viral, with EBV and CMV leading. A particularly glucocorticoid-responsive variant occurs with systemic autoimmune disease, where it is called macrophage activation syndrome — in SLE, adult-onset Still’s disease and systemic JIA.
Checkpoint-inhibitor-associated HLH is rare but increasingly recognised: a WHO database recently held 681 such reports — 231 for nivolumab, 177 for pembrolizumab, 169 for ipilimumab, 90 for atezolizumab and 12 for durvalumab. The driving mechanism is disinhibition of the costimulatory molecules that normally restrain T-cell activation; that disinhibition is the desired therapeutic effect, but it can tip into inappropriate release of IL-6, TNF-α and interferon-γ.
Management beyond stopping the drug and giving high-dose glucocorticoids includes anakinra (IL-1 receptor antagonist) and ruxolitinib (JAK1/2 inhibitor), with low-dose etoposide considered in refractory cases. Given the interferon-γ elevation, some have proposed a role for emapalumab, an anti-interferon-γ antibody effective in an open-label study of children with primary HLH.
The wider lesson is that systemic inflammatory syndromes deserve a place in the differential for fever in a patient with cancer — and that this remains an under-recognised manifestation of immunotherapy toxicity.