Pyrexia of Unknown Origin: The Evidence Behind the Pathway

A 45-year-old farmer from Shivamogga district has had fever for five weeks. He has lost 6 kg. He sweats at night. He has been seen three times in the OPD, and he has had three courses of antibiotics, a malaria test, a Widal test and two blood counts. His chest X-ray is normal. His file now says “PUO”, and the intern asks, “Sir, what else should we order?” This is an illustrative case, not a real patient.

That question is the problem. Pyrexia of unknown origin is not solved by ordering more. It is solved by confirming the fever, following the clues, and getting tissue from the right place. By the end of this article you will know why the definition exists and why tuberculosis comes first in India. You will know why the next step for this man is probably a biopsy rather than a blood test, and how to start an empirical trial without losing your way.

1. What the definition is for

The definition has a job. The threshold exists to stop the expensive, low-yield investigation of a fever that would have declared itself if given a fortnight. Most fevers are short, and most short fevers are infections that either settle or show themselves within two weeks.

Petersdorf and Beeson set the first definition in 1961: fever above 38.3 C, which is 101 F on the charts most Indian wards actually use, on several occasions, lasting more than three weeks, and undiagnosed after a week of investigation in hospital. Durack and Street revised it in 1991. The week in hospital became three outpatient visits or three days of inpatient investigation, and they added the nosocomial, neutropenic and HIV-associated categories. The tool sorts every case into one of those four categories, because each has a different list of causes.

Our farmer qualifies: more than three weeks of fever, three outpatient visits, no diagnosis. A fever of less than two weeks belongs in the acute undifferentiated febrile illness pathway instead, which is built for the tropical infections that dominate that window.

2. Confirm the fever before you investigate it

“But he says he has fever every evening.” He may. A surprising proportion of referred PUO is not PUO: a normal circadian variation misread as fever, a reading taken only at home on an unreliable thermometer, or no documented fever at all. So before any imaging, ask for the temperature charted by a health worker, four hourly, alongside the pulse.

The pulse earns its place on that chart. A fever above 102 F (39 C) with a pulse under 90 raises relative bradycardia, which points towards enteric fever among other causes. A fever with no rise in pulse and no rise in acute phase markers raises the possibility that the reading itself is not genuine. Approach that possibility with care and without accusation.

3. Why tuberculosis comes first in India

Here is what the Indian series show. Indian series consistently place infection as the largest group in classic PUO, and tuberculosis as the commonest single infection within it, much of it extrapulmonary and much of it with a normal chest radiograph.

Read the last part again: much of it with a normal chest radiograph. This is where residents most often go wrong, and it is an honest mistake. A normal chest X-ray feels as if it has ruled tuberculosis out. In PUO it has not. The tuberculosis is often in a node, the pleura, the peritoneum, the marrow or the liver.

So the search is for a site to sample rather than for another blood test. Lymph node, pleura, peritoneum, marrow and liver are all more informative than another round of serology. And the tissue must go for GeneXpert and mycobacterial culture as well as histopathology. “Is a granuloma on histology not enough?” No. A granuloma alone does not distinguish tuberculosis from sarcoidosis, fungal infection or lymphoma.

4. Follow the clues, not the checklist

Ordering every test in a long list is slower and less productive than following the clues that are present. The tool’s Differential tab is built that way: for each cause, what should raise it, and the test that settles it. A few rows show the idea.

“And if there are no clues?” Then repeat the history and the examination, not the tests. Clues appear over time, and the examination that was normal in week two is often abnormal in week four. A node that was not there last week can be the finding that settles the case.

Imaging follows the same order of sense. Ultrasound of the abdomen is cheap, available and reasonable first. A contrast-enhanced CT of the chest and abdomen is the workhorse of the second stage, and finds most occult abscesses, lymphadenopathy and solid tumours. FDG PET-CT is useful where available and affordable, particularly for large vessel vasculitis and for finding a site to biopsy, but it is not available in most Indian district settings, and a workup should not stall waiting for it.

5. A trial of treatment needs a plan

“Why not just start antitubercular treatment and see?” It is widely done in India, where the pretest probability of tuberculosis is high, and it may be the only practical option when tissue cannot be obtained. But it has two costs. Both lymphoma and Still’s disease can settle transiently on any of these agents, so a fall in temperature does not confirm the diagnosis. And a full course commits the patient to months of hepatotoxic treatment.

Steroids carry their own warning: a striking response in giant cell arteritis or Still’s disease, a partial and temporary one in lymphoma, and the risk that tuberculosis disseminates. So the tool asks you to record what is being treated, what response would count as a success, by what date it will be judged, and what will happen if it does not come. Where tissue can still be obtained, obtaining it before starting is almost always the more informative order.

You may be asked in an examination about the naproxen test: a short course said to separate neoplastic fever, which settles, from infective fever, which does not. The tool keeps the one thing you need at the bedside, that it is a weak discriminator in modern series and should not be used to rule anything in or out. Why does the tool mention it at all? In its own words: It appears here because it is still asked about in examinations and still occasionally used on the wards, not because it should decide management.

Back to the farmer

So what does the intern order? Not another Widal. First, a charted temperature with the pulse, to confirm the fever. Then an HIV test, which belongs in every case of PUO, whether or not risk factors are volunteered. Then look again for a site: an ultrasound of the abdomen, and a careful search for nodes. In our illustrative case the ultrasound shows enlarged nodes in the abdomen, and a CT-guided biopsy is sent for GeneXpert, mycobacterial culture and histopathology. The answer comes from the tissue, not from a fourth course of antibiotics.

Try this

A 30-year-old man has had fever for four weeks after two courses of antibiotics from a local clinic. He has vague abdominal pain. On the charted readings his temperature is 103 F with a pulse of 84. What does that pulse suggest, and what test settles it? Illustrative case.

Answer. A fever above 102 F with a pulse under 90 is relative bradycardia. With prior antibiotics and abdominal symptoms, the tool’s Differential tab points to enteric fever, relapsing or partially treated. The test that settles it is a blood culture, and a bone marrow culture if he is already on antibiotics.

Take-home points

• Confirm the fever with a charted temperature and pulse before you investigate it.
• In India, think of tuberculosis first, and do not let a normal chest X-ray rule it out.
• Look for a site to sample; send tissue for GeneXpert and culture as well as histopathology.
• Follow the clues; when there are none, repeat the history and examination rather than the tests.
• Before an empirical trial, write down what you are treating, what counts as success and by when.

Declaration: the author built and maintains the tool. It is free to use. The figures in this article are those the tool carried on 5 November 2026, each from the source the tool names. The illustrative cases follow the tool’s rules; check the tool’s result for a real patient.

References

1. Petersdorf RG, Beeson PB. Fever of unexplained origin: report on 100 cases. Medicine (Baltimore). 1961;40:1-30.
2. Durack DT, Street AC. Fever of unknown origin: reexamined and redefined. Curr Clin Top Infect Dis. 1991;11:35-51.
3. Mulders-Manders C, Simon A, Bleeker-Rovers C. Fever of unknown origin. Clin Med (Lond). 2015;15(3):280-284.
4. Haidar G, Singh N. Fever of Unknown Origin. N Engl J Med. 2022;386(5):463-477.
5. Bleeker-Rovers CP, Vos FJ, de Kleijn EMHA, et al. A prospective multicenter study on fever of unknown origin: the yield of a structured diagnostic protocol. Medicine (Baltimore). 2007;86(1):26-38.
6. Kejariwal D, Sarkar N, Chakraborti SK, Agarwal V, Roy S. Pyrexia of unknown origin: a prospective study of 100 cases. J Postgrad Med. 2001;47(2):104-107.
7. Pannu AK, Golla R, Kumari S, Suri V, Gupta P, Kumar R. Aetiology of pyrexia of unknown origin in north India. Trop Doct. 2021;51(1):34-40.
8. Sharma BK, Kumari S, Varma SC, Sagar S, Singh S. Prolonged undiagnosed fever in northern India. Trop Geogr Med. 1992;44(1-2):32-36.
9. Central TB Division, Ministry of Health and Family Welfare, Government of India. National Tuberculosis Elimination Programme: Training Modules and Technical and Operational Guidelines for TB Control in India. New Delhi: MoHFW.
10. Fautrel B, Zing E, Golmard JL, et al. Proposal for a new set of classification criteria for adult-onset Still disease. Medicine (Baltimore). 2002;81(3):194-200.
11. Fardet L, Galicier L, Lambotte O, et al. Development and validation of the HScore, a score for the diagnosis of reactive haemophagocytic syndrome. Arthritis Rheumatol. 2014;66(9):2613-2620.
12. Li AY, Adams-Huet B, Cutrell JB, et al. Diagnostic yield of FDG PET-CT in fever of unknown origin. Clin Infect Dis. 2021;73(11):e4374-e4382.