ISSN: 0443-511
e-ISSN: 2448-5667
Herramientas del artículo
Envíe este artículo por correo electrónico (Inicie sesión)
Enviar un correo electrónico al autor/a (Inicie sesión)
Tamaño de fuente

Open Journal Systems

Fever of unknown origin, comparing two series with 26 years of difference

How to cite this article: Casarrubias-Ramírez M, Alfaro-Mejía JA, De Santiago-Leaños J, Mendoza-Álvarez SA, Pineda-Galindo LF, Vera-Lastra OL. Fever of unknown origin, comparing two series with 26 years of difference. Rev Med Inst Mex Seguro Soc. 2015;53 Supl 1:S6-S17.



Received: October 22nd 2014

Accepted: March 6th 2015

Fever of unknown origin, comparing two series with 26 years of difference

Moisés Casarrubias-Ramírez,a José Alfredo Alfaro-Mejía,b Juan De Santiago-Leaños,c Sergio Alberto Mendoza-Álvarez,d Luis Francisco Pineda-Galindo,e Olga Lidia Vera-Lastraf

a,c,d,e,fDivisión de Estudios de Posgrado, Facultad de Medicina, Universidad Nacional Autónoma de México.

a-fDepartamento de Medicina Interna, Hospital de Especialidades Centro Médico Nacional La Raza. Instituto Mexicano del Seguro Social.

Distrito Federal, México

Communication with: Moisés Casarrubias-Ramírez.

Telephone: (55) 5689 5187


Background: Fever of unknown origin (FUO) remains a syndrome with difficult approach and changing spectrum. Our aim was to compare two series of FUO patients seen at the Department of Internal Medicine, Hospital de Especialidades Centro Medico La Raza, Mexico City.

Methods: Data from FUO series from 1979-87 were compared with those from 2004-14 series. We analyzed demographic data, final diagnoses, and diagnostic tests used. We report median and range for numerical variables and frequencies for nominal data, bivariate analysis was done with chi-square or Fisher´s test as needed using SPSS version 17.0 for MAC and open-epi version 3.7.

Results: One hundred twenty seven patients were included in the 1979-87 series and 118 in the 2004-14 series. There were more non-infectious inflammatory diseases (p = 0.0004) and less infectious diseases (p = 0.024) in the 2004-14 series. We observed no significant differences in neoplastic diseases and undiagnosed cases between the two series. Laboratory tests and their diagnostic utility were similar in both series, but image studies were less useful in the 2004-14 series. Biopsy and laparotomy remained as frequent and useful tools in both series.

Conclusions: The recent series had more non-infectious inflammatory diseases and less infectious causes of FUO. Invasive studies remain as useful diagnostic aids in a significant number of cases.

Keywords: Fever of unknown origin, Non-infectious inflammatory diseases, Infectious diseases, Neoplastic diseases.

Fever of unknown origin (FUO) is a syndrome characterized by a clinical spectrum of patients coming to the medical office with fever as the only or predominant manifestation, of considerable magnitude (> 38.3 degrees), prolonged (> 3 weeks), and in whom the results of initial clinical evaluation and paraclinical studies have not identified its cause.2 To simplify the reporting of cases of this syndrome, they have been grouped into five categories: infectious, neoplastic, non-infectious inflammatory diseases (NIID), miscellaneous, and unknown origin.2 Due to the great diversity of causal diseases, this syndrome represents a diagnostic challenge for the physician and a significant consumption of resources for health institutions.

The criteria for classifying FUO have evolved since Petersdorf proposed the following in 1961: fever greater than 38.3 °C on several occasions, for more than three weeks, and uncertain diagnosis after a week of study in hospitalization;3 these criteria have been modified by changes in the epidemiological profile of febrile diseases, the availability of new resources for diagnosis, the emergence of new diseases, and changes in styles of medical practice in different health systems.

Durak and Street4 proposed eliminating the need for a week of hospitalization and replacing it with a hospital stay of only 3 days, or 3 subsequent outpatient visits without reaching an etiologic diagnosis, reflecting a shift towards outpatient medicine, and increased availability of resources to study patients in less time. Finally, Knockaert and other authors2,5 have proposed a minimal and more or less uniform list of negative diagnostic studies to classify a patient as having FUO.1 These changes have resulted in the selection of patients with a more difficult approach, modifying the relative distribution of the different causal diseases6,7 and increasing the proportion of cases without etiological diagnosis.8,9 This has also caused changes in the diagnostic tests used.9,10

Knowing local and national epidemiology as to the causes of the syndrome, as well as useful diagnostic studies, is of great help in the management of these patients. This justifies the periodic analysis of local caseloads.1 For this reason we decided to do a comparative analysis of two studies of patients with FUO studied with 26 years of difference in the Hospital de Especialidades of the Centro Médico Nacional La Raza of the Instituto Mexicano del Seguro Social.


In 1988, Frati et al.1 published a descriptive study of FUO, retrospectively analyzing records (n = 97) and clinical summaries (n = 30) of 127 patients hospitalized in the Departamento de Medicina Interna of the Hospital de Especialidades of the Centro Médico La Raza between May 1979 and April 1987, reporting demographic and clinical data, causes of the syndrome, and utility of various diagnostic tests. Data for the 1979-1987 study were taken from this publication.

The Departamento de Medicina Interna at this hospital has a registered database of patients with FUO, which was built in two stages. Initially data were retrospectively collected from medical records of patients hospitalized between the last quarter of 2003 through December 2007. Starting in January 2008, cases were recorded prospectively, making a total of 118 cases by May 31st 2014. Data for the 2004-2014 study data were taken from this database.

The selection criteria for each study are summarized in Table I and reflect the prevailing criteria in each period. Considering that this is a study of secondary data (a publication and a database), the analysis was restricted to the data of the previous publication that could be subject to a comparative analysis with the current database. Thus, the analysis compared incidence of the syndrome as cause of admission, distribution by age and sex, length of hospital stay, relative frequency of causal diseases, and diagnostic tests reported as decisive for the final diagnosis, which were classified as laboratory studies, imaging, biopsies, and protocolized laparotomy. Because the 1979-1987 study lacks individual data on age and length of hospital stay, comparative statistical analysis of these variables was not made. The proportion of the different causal categories, specific disease entities, useful diagnostic tests, and the proportion of deaths were compared with chi-squared or Fisher's exact test, according to the number of analyzable subjects and expected values ​​for each variable.

Table I Selection criteria for the two studies of fever of undetermined origin in the study
1979-87 Study (Petersdorf) 2004-14 Study (Knockaert)
1. Fever ≥ 3 weeks duration 1 Fever ≥ 3 weeks duration
2. Temperature ≥ 38 degrees confirmed in the hospital on at least two occasions 2 Temperature ≥ 38.3 degrees confirmed in the hospital on at least two occasions
3. No diagnostic orientation after:
  • Clinical history
3. No diagnostic orientation after:
  • Clinical history
  • Hematology and erythrocyte sedimentation
  • Routine blood tests (hematology, blood chemistry, hepatic profile, electrolytes, and muscle enzymes)
  • Liver function tests
  • Febrile reactions
  • General urine test
  • General urine test
  • PA chest X-ray
  • Blood and urine cultures (at least 2)
  • Simple abdominal plate
  • PA chest X-ray
  • Abdominal-pelvic ultrasound
  • Antinuclear, anti-neutrophil cytoplasmic, and rheumatoid factor antibodies
  • Infectious serology according to local epidemiology (febrile reactions, IgM serology for Cytomegalovirus and Epstein Barr virus, viral hepatitis panel).

The descriptive analysis of the 2004-2014 study was performed using SPSS version 17.0 for MAC, and comparative analysis between the two studies for data summary was performed in open-epi program version 3.7. p < 0.05 was considered statistically significant, and all tests were two-tailed.


Sociodemographic and clinical variables

The 1979-1987 study included 127 cases that accounted for 1.06% of the 11,896 admissions in the department during that period, for an average of 15.8 cases per year. A total of 65 cases (51.2%) were females, and age ranged between 16 and 80 years, with a median of 37. Final diagnoses were grouped into: infections, 53 cases (41.7%); NIID, 23 (18.1%); neoplasia, 22 (17.3%); miscellaneous, 9 (7.1%), and fever of undetermined origin, 20 cases (15.7%) (Table II). The time from admission until time of definitive diagnosis ranged from 4 to 300 days, with a median of 15 days. No significant differences were detected in the clinical manifestations of the different causal categories, only a faster weight loss was found in the group of patients with neoplastic diseases (2.3 kg/month).

Table II Distribution of diagnostic categories for studies of fever of undetermined origin
Diagnostic category 1979-87 Study 2004-14 Study p*
Infections 53(41.7%) 33(27.9%) 0.024
Neoplasms 22(17.3%) 23 (19.5%) 0.662
Non-infectious inflammatory diseases 23(18.1%) 40(33.9%) 0.004
Miscellaneous causes 9 (7.1%) 3(2.5%) 0.094
Fever of undetermined origin 20(15.7%) 19(16.1%) 0.930
Total 127 118
Chi-squared in all comparisons.

The 2004-14 study included 118 cases that accounted for 1.3% of the 8982 admissions in that period, for an average of 11.8 cases per year. Of these, 71 cases (60.2%) were male, and age ranged from 16 to 87 years, with a median of 40. The final diagnoses included 33 cases of infectious diseases (27.9%); 40 of NIID (33.9%); 23 patients with neoplasms (19.5%); 3 cases of miscellaneous diseases (2.5%); and 19 cases of fever of undetermined origin (16.1%). These results show a significant increase in the frequency of NIID (p = 0.004) at the expense of a decrease in the number of infections (p = 0.024). No significant difference was observed in the proportion of patients with fever of unknown origin at the end of the study protocol. The time from admission to time of diagnosis ranged from 4-180 days with a median of 22, representing an increase of one week of study time relative to the previous study. A higher frequency of weight loss was also found in the neoplastic category (OR 5.3, 95% CI 1.67-16.68, p = 0.002), although the rate of weight loss could not be estimated because there was no quantitative record of it in the database.

Final diagnoses by categories


Infectious causes of both studies are summarized in Table III. For purposes of analysis, they were classified arbitrarily into systemic and localized. In both studies systemic infections were much more frequent. For the 1979-1987 study tuberculosis was the most common cause (13 cases), all of which were extrapulmonary, followed by endocarditis (7 cases), brucellosis and salmonellosis (6 cases each), amoebic liver abscess (5 cases), malaria, and abdominal abscesses (3 cases each). For the 2004-2014 study, the most common infection was also tuberculosis (17 cases), eleven of them with disseminated TB (two or more documented locations or miliary tuberculosis) and only 3 with pulmonary location; followed by septicemia of unknown origin (3 cases), cholangitis, orchitis, and abdominal abscesses (2 cases each). In both studies there is a predominance of systemic infections in general and tuberculosis in particular, however there is a clear reduction in cases of endocarditis and a virtual disappearance of brucellosis and malaria. The proportion of tuberculosis cases increased in the 2004-14 study with 68 versus 33.3% of systemic infections (p = 0.013). Abdominal abscesses were a common cause of FUO secondary to localized infection in both groups of patients.

Table III Distribution of infectious causes for fever of undetermined origin by study
Systemic infections Localized infections
Study: 1979-87 2004-14 Study: 1979-87 2004-14
n= 39 n= 25 n= 13 * n= 8
Infection No (%) No (%) p* Infection No (%) No (%) p
Tuberculosis 13(33.3) 17(68.0) .013 Amoebic hepatic A 5(38.4) 0(0) .233
Endocarditis 7(17.9) 1(4.0) .201 Abdominal abscess 3(23.0) 2(25.0) NS
Brucellosis 6(15.4) 0(0.0) .175 Catheter colonization 1(7.7) NS
Salmonellosis 6(15.4) 1(4.0) .312 Bronchiectasis 1(7.7) 0(0.0) NS
Malaria 3(7.7) 0(0.0) .706 Sinusitis 1(7.7) 0(0.0) NS
Hepatitis B 1(2.6) 0(0.0) NS Proctitis 1(7.7) 0(0.0) NS
Septicemia 1(2.6) 3(12.0) .322 Meningitis 1(7.7) 0(0.0) NS
Toxoplasmosis 1(2.6) 1(4.0) NS Orchitis-epididymitis 0(0) 1(12.5) .875
Cytomegalovirus 0(0.0) 1(4.0) NS Cholangitis 0(0) 2(25.0) .875
Chagas disease 0(0.0) 1(4.0) NS Urinary tract inf. 0(0) 2(25.0) NS
HIV infection* 1(2.6) 0(0.0) NS 1(12.5)
* All comparisons with Fisher's test.The values reported not significant (ns) correspond to calculations ofp> 0.999 with open-epi program, other values are reported with exact probability.
* 1979-87 study did not report the diagnosis of one of their localized infections.


Neoplastic causes are summarized in Table IV. The relative frequency of this category had little variation between the two studies, 17.3 versus 19.5% (p = 0.661). Of the 22 cases in the 1979 to 1987 study, 16 were hematologic neoplasms (72.7%), lymphomas being the most common with 11 cases (50%), Hodgkin's disease predominating in 9 of them. Only 6 patients had carcinomas (27.3%), the most common being metastatic cancer of unknown primary with 3 cases (13.6%), and breast cancer with two (9%).

Table IV Distribution of neoplastic causes of fever of undetermined origin by study
Hematologic Neoplasms Carcinomas
Study: 1979-87 2004-14 Study: 1979-87 2004-14
n= 16 n= 17 n= 6 n= 6
Neoplasia No (%) No (%) p* Neoplasia No (%) No (%) p*
Hodgkin's disease 9(56.2) 4(23.5) .116 Ca unknown primary 3(50.0) 3(50.0) NS
Non-Hodgkin's L 2(12.5) 7(41.2) .142 Breast 2(33.3) 0(0) NS
Multiple myeloma 2(12.5) 1(5.9) .953 Vesicle 1(16.6) 0(0) NS
Macroglobulinemia 1(6.2) 0(0) NS Prostate 0(0) 1(16.6) NS
Leukemias 1(6.2) 1(5.9) NS Peritoneal  0(0) 1(16.6) NS
Histiocytosis 1(6.2) 1(5.9) NS Carcinomatosis 0(0) 1(16.6) NS
Myeloid metaplasia 0(0) 2(11.8) NS Retroperitoneal tumor
Myelodysplastic s. 0(0) 1(5.9) NS
* All comparisons with Fisher's test.The values reported not significant (ns) correspond to calculations of p > 0.999 open-epi program, other values are reported with exact probability.

Of the 23 cases in the 2004-2014 study, 17 had hematological malignancies (73.9%) with 11 lymphomas (47.8%), but with an increase in the relative frequency of Non-Hodgkin’s lymphoma with 7 cases (41.2%), which was not significant (p = 0.142). A total of 6 patients had carcinomas and half of them had metastatic cancer of unknown primary, not considering other patients with peritoneal carcinomatosis (which is also a malignant tumor with unidentified primary). The distribution of diseases in this category showed minor variations highlighting the inversion of the types of lymphoma diagnosed.

Noninfectious inflammatory diseases

The causes of this category are summarized in Table V. A significant increase of these cases was observed in the 2004-2014 study (p = 0.004), and although systemic lupus erythematosus predominates in both as the most common cause with 11 and 18 cases respectively (47.8 versus 45.0%), this highlights the relative increase in cases of Still's disease in the more recent caseload (n = 11), although the difference was not significant (17.4 versus 30%, p = 0.423). There are no substantial differences in the other diseases in this category. One case from the 1979-1987 study was reclassified as their final diagnosis was autoimmune hepatitis, and the disease is now classified as non-infectious inflammatory and not as miscellaneous.

Table V Distribution of non-infectious inflammatory causes for fever of undetermined origin by study
Disease Series 1979-87
n= 127
No (%)
Series 2004-14
n= 118
No (%)
Systemic lupus erythematosus 11(47.8) 18(45.0) p=ns
Still's disease 4(17.4) 12(30.0) p=.423
Dermatomyositis 2(8.7) 0(0) p=ns
Polyarteritis nodosa 2(8.7) 1(2.5) p=.598
Other vasculitis 2(8.7) 3(7.5) p=ns
Polymyalgia rheumatica 1(4.3) 0(0.0) p=.989
Overlap syndrome 0(0.0) 1(2.5) p=ns
Antiphospholipid syndrome (DVT) 0(0.0) 1(2.5) p=ns
Autoimmune hepatitis 1(4.3) 0(0.0) p=.989
Sarcoidosis 0(0.0) 1(2.5) p=ns
Wegener's granulomatosis 0(0.0) 1(2.5) p=ns
Erythema nodosum 0(0.0) 1(2.5) p=ns
Total 23(100) 40(100)
* All comparisons with Fisher's test.The values reported not significant (ns) correspond to calculations ofp> 0.999 open-epi program, other values are reported with exact probability.


This category was reduced from 9 cases in the 1979-1987 study to just 3 cases in 2004 to 2014. The study by Frati et al.1 reported 4 cases of factitious fever, two of liver cirrhosis, two patients with Caroli disease, as well as one case of hemolytic anemia and another of autoimmune pernicious anemia. In the current study no FUO cases were reported in patients with liver cirrhosis and no cases of Caroli disease were observed, nor of factitious fever; the latter because all cases in the 2004-14 study required the fever to be confirmed by a temperature measurement monitored by health personnel. However, two cases of granulomatous hepatitis were reported, which remains a diagnosis reported in the FUO studies, but not a definitive diagnosis. Previously these cases received therapeutic trial with anti-tuberculosis drugs and if they improved they were reclassified as tuberculosis. Currently, if a specific diagnosis cannot be established and clinical conditions are stable, they are followed without treatment and fever usually subsides spontaneously as in these two cases. The other case in the 2004-14 study in this category was a patient with inflammatory pseudotumor.

Cases with indeterminate diagnosis

A total of 20 cases with unknown causal diagnosis were presented in the 1979-87 study and 19 in the 2004-14 study (15.7 versus 16.1%, p = 0.939). In both studies the patients in this category had a self-limiting fever that resolved spontaneously in 17 cases of the 1979-1987 study and in 18 cases in the 2004-14 study (84.2 versus 95%, p = 0.268). Two cases of the 1979-1987 study received therapeutic trial with anti-tuberculosis drugs and the fever also resolved, while another patient died undiagnosed. In the 2004-14 study, one patient was lost to follow-up so their evolution is unknown.

Diagnostic tests

Laboratory tests: Laboratory tests (which are summarized in Table VI) were decisive for the diagnosis. In both studies antinuclear antibodies were the laboratory test that allowed the most cases to be diagnosed, this due to the high frequency of systemic lupus erythematosus (SLE) in both studies. This test was positive only in patients with collagen diseases in the 1979-1987 study, but the authors did not report all patients tested with this test. In the 2004-14 study, the test had a higher number of false positives (all with figures ≥ 1/160) since it was performed as a screening in almost all patients to be classified as FUO. In both patient groups cultures, stains, and serology for infectious agents were useful, but the smears tended to be replaced by molecular biology techniques such as PCR which is more sensitive and, especially, more specific.

Table VI Decisive laboratory and imaging tests for diagnosis of fever of undetermined origin by study
Diagnostic test 1979-87 Study
n= 127
No (%)
2004-14 Study
n= 118
No (%)
Immunology (ANAS, ANCAS, FR Acl) 12(8.7) 20(14.4) .120
Microbiology (cultures and stains) 2(16.5) 7(5.9) .014
Serology of infectious agents 7(5.5) 5(3.9) .241
Protein electrophoresis 2(1.6) 2(1.6) p= ns
Hematology 1(0.8) 0(0.0) p= ns
PCR forMycobacterium tuberculosis 0(0.0) 5(3.9) p=.894
Total 43(33.8) 38(32.2) p=.783
Abdominal tomography 10(7.8) 3(2.5) .111
Abdominal ultrasound  9(7.1) 2(1.7) .073
Conventional X-rays 8(6.3) 4(3.4) .450
Echocardiography 6(4.7) 1(0.8) 145
Gammography 5(3.9) 1(0.8) .249
Venous Doppler 0(0.0) 1(0.8) p= ns
Total 38(29.9) 12(10.2) .0001
* Comparisons between individual tests with Fisher's exact test and comparison of data grouped by study type with chi-squared.The values reported not significant (ns) correspond to calculations ofp> 0.999 open-epi program, other values are reported with exact probability.

Imaging tests: the most useful imaging tests for diagnosis in both studies are summarized in Table VI. One can see a clear decrease in imaging studies (and particularly computerized tomography and abdominal ultrasound) as conclusive evidence for the diagnosis of the cause of FUO, between 1979-1987 and 2004-2014’s studies. Imaging tests were decisive in 38/127 cases for the 1979-1987 study and only 12 of 118 cases for the 2004-2014 study (29.9 versus 10.2%), p = 0.0001). This is related to a decrease in cases of amoebic liver abscess and abdominal abscesses between the two studies, and the fact that a negative ultrasound is a screening test now required to classify a patient with FUO. Other simple radiographic techniques have fallen into disuse in the approach to these patients.

Endoscopic studies: Only 3 cases between the two studies were diagnosed with endoscopic studies.

  • Biopsies: Biopsies allowed diagnosis in 25 cases in the 1979-87 study and 29 patients in the 2004-14 study (19.6 versus 24.5%, p = 0.356). It is unknown how many biopsies were performed in the patients of the 1979-87 study, as only those cases that were diagnostic (Table VII) were reported. In the 2004-2014 study, at least one biopsy was performed on 68 patients (57.6%). In both studies the most useful biopsies were of lymph node, liver, and bone marrow. Skin biopsy allowed a greater number of diagnoses in the 2004-2014 study. No differences were observed in either the trend of execution, or the diagnostic utility of these invasive procedures.  

  • Table VII Decisive biopsies and laparotomies for diagnosis fever of undetermined origin by study
    Biopsy 1979-87 Study
    n= 127
    No (%)
    2004-14 Study
    n= 118
    No (%)
    Lymph node 8(6.3) 7(5.9) p= ns
    Liver 6(4.7) 10(8.5) .353
    Bone marrow 4(3.1) 5(4.2) .908
    Muscle 2(1.6) 0(0.0) .814
    Skin 2(1.6) 5(4.2) .388
    Kidney 1(0.8) 1(0.8) p= ns
    Breast 2(1.6) 0(0) .814
    Prostate 0(0.0) 1(0) p= ns
    Laparotomy 10(7.8) 13(11.0) .532
    Total 35(27.5) 42(35.6) .176
    * Comparisons between individual tests with Fisher's exact test and comparison of data grouped by study type with chi-squared. The values reported not significant (ns) correspond to calculations of p > 0.999 open-epi program, other values are reported with exact probability.

  • Exploratory laparotomy: This was done in 16/127 in cases of the 1979-87 study and 20/118 patients in the 2004-2014 study (12.6 versus 16.9%, p = 0.336), and was diagnostic in 10 cases in the previous study and 13 in the current study (62.5 versus 65%, p = 0.876). Contrary to expectations, no reduction in the use of this procedure was observed and, on the other hand, a diagnostic yield greater than 60% was maintained. In over half the cases of both studies, the diagnosis obtained by laparotomy was for cases of lymphoma and tuberculosis, and in fewer cases for intra-abdominal pathologies.

The duration of disease of patients in the 1979-1987 study is unknown, since only the number of deaths was reported. A total of nine patients died (7.1%), three with infections, three non-infectious inflammatory diseases, two with neoplasms, and one with indeterminate diagnosis.

In the 2004-14 study, 47 patients were cured (39.8%), 49 patients improved (41.5%), 12 patients died, and 10 patients were discharged alive but their outcome at six months remains unknown, 9 of them because they were referred to other services or hospitals for further treatment, and the other because they were lost to follow-up. There was no difference in deaths (7.1 versus 10.2%, p = 0.389). The 2004-14 study showed a predominance of neoplastic diseases in patients who died (8 cases), which did not have statistical significance due to the low number of registered deaths (22.2 versus 66.6%, p = 0.112).


Frequency of causes of FUO is influenced by epidemiological aspects, case referral filters, and the resources available for studying patients. For this reason there has been a transition modifying the clinical spectrum of patients with this syndrome.

The results of Frati1 are consistent with most studies contemporary to it, with the greatest frequency of infections, followed by neoplastic diseases, NIID third, and fewer cases with undetermined diagnosis, usually less than 20%. Such findings were reported both in Mexico11-13 and abroad.14-16

Comparison of these two studies shows a reduction in infectious causes and an increase in patients with NIID, with a reduction in neoplastic diseases. This trend has been reported by Knockaert and De Kleijn since the 90's, when both reported 25% for infectious diseases and a figure of 24-26% for NIID, and it became more evident in studies of Vanderschueren and Bleeker-Rovers, published in 2003 and 2007 respectively, in which NIID was the main cause and infections fell below 20% of cases.8,9 Efstathiou,17 in Greece, also reported NIID as the first category with 33% of cases, just above infections with 30%. This distribution, however, is not uniform, and there are recent studies in other regions of Europe, such as those reported by Colpan18 and Baicus19 in Turkey and Romania, which continue to show a predominance of infections like Liu20 in Taiwan and Naito21 in Japan.

Hayakawa et al.22 conducted a review of 26 studies of FUO published between 2000 and 2011 that included at least 50 patients, and the sum of these studies shows that infections remain the leading cause of FUO with 36%, followed by NIID with 20.9%, and a rising ratio of undiagnosed cases with 23.7%. In general, studies from developing countries, or those performed in community hospitals or referral hospitals receiving patients from rural areas, continue to show a predominance of infections, while studies from developed countries, university hospitals, or urban centers have shown a turn with increased NIID. The only recent study published in Mexico by Arce-Salinas23 in a hospital in Mexico City, reported 42% of infections and 26.6% NIID with only 11.1% of undiagnosed cases, however, the small number of patients may have caused an unrepresentative sample. The population treated in our hospital is made up of salaried workers who live predominantly in urban areas, and who are referred to general hospitals that represent a previous level of care where a significant proportion of cases are resolved. This selection filter could cause an insignificant representation of infections as causes of FUO in our sample. Another cause may be the fact that our latest study excluded all patients with HIV infection, considering it a separate category according to the classification proposed by Durack and Street, while others do include those cases that did not have previous diagnosis of infection when FUO began. This may explain the difference, since HIV infection is one of the most common infectious causes reported by these same authors.20,23

Specific infections have also shown variations and are more influenced by local epidemiology, however most studies, regardless of origin, continue to report tuberculosis as the most common infection, particularly extrapulmonary,18,20 a finding shared by the two studies in our hospital, despite their distance in time. Other frequently reported infections include endocarditis,17,21 salmonellosis, and brucellosis,24 which, although common in the 1979-87 study, were very rare in the current study. On the other hand, there is a global trend towards reduced localized infections, particularly intracavitary abscesses, a finding also documented in our latest study.

As for NIID, SLE remains a common cause,9,17,18 with increased Still's disease,9,16,17 data shared by our caseload study, while other common diseases in older populations such as polymyalgia rheumatica and arteritis temporalis19,17 which only represented one case of our two studies.

The number of neoplasms remains relatively constant as the third causal category with a predominance of hematological neoplasms in all studies, varying only in the relative distribution between Hodgkin's disease and non-Hodgkin’s lymphoma, the latter being the most common, in other studies and in ours.9,17,18 

Undiagnosed patients have increased from 7% in Petersdorf’s initial study3 up to 51% in the study from Bleeker Rovers,9 and the global trend is towards a variable increase in these cases.8,21,22 This finding may have different explanations: a greater number of cases diagnosed by autopsy in the initial studies (which would increase cases not diagnosed in life by Petersdorf to 16%); change in diagnostic criteria for FUO selecting the patients most difficult to diagnose; the abandonment of "soft criteria" for the diagnosis of some diseases that now require microbiological, histological, and molecular confirmation; and the increasingly frequent use of "therapeutic testing" as a diagnostic criterion (as in many patients with FUO the fever limits itself). The proportion of indeterminate cases also increased in the first Mexican studies, from 5.45% in the study by Lascuraín11 in the 70’s, to 13.3-15.6% in the studies in the 80’s.1,12,13 However, this proportion has remained relatively constant since then, ranging from 11.1% reported by Arce-Salinas23 and 16.1% found by us, lower figures but not very far from the 23.7% reported in Hayakawa’s review22 summarizing the most representative studies of this decade. Despite these variations, the published works are consistent in that most patients with indeterminate diagnosis who are clinically stable should be followed with symptomatic treatment, because the fever will subside spontaneously in most of them, as happened in our patients.25

It was not possible to compare sensitivity, specificity, and predictive values ​​of diagnostic tests used between the two studies, as Frati’s publication reported only useful studies (referred to as decisive), and not all studies, which precludes us knowing the false positive and negative results. Indeed, studies evaluating the utility of diagnostic tests in FUO, have uncertain estimates of these parameters, because only those tests are considered positive that "contributed" to the diagnosis or that represented positive "diagnostic clues." Therefore, we only compare the proportion of patients who were diagnosed with the different tests.

As for decisive laboratory tests, it was found that the more recent study had fewer patients whose cultures and stains were useful (p = 0.014), while a larger share was diagnosed using various immunological studies (mainly ANA and ANCAS), which reflects the change in the distribution of causal diseases. The utility of serology for infectious agents was low in both studies, and these findings correspond with those reported by other authors.8,9,26

Imaging studies were significantly less useful in the 2004-14 study, with only 10% of all cases diagnosed by all imaging studies combined, and in return the trend of performing invasive tests was maintained, such as puncture and excision biopsies, and protocolized laparotomy. This coincides with other authors, such as Vanderschueren8 who reported that biopsies were decisive for diagnosis in 24.5% of cases versus only 16.7% of all imaging studies combined, and only 6.3% for computerized tomography and gammography respectively. Mete27 reported similar findings in Turkey, where diagnosis required invasive studies in 49% of cases, and both studies report a range of causes similar to those of our 2004-14 study, with NIID as the most frequent cause. Hayakawa22 reported that CAT has a percentage of utility of 17-20% of cases and up to 21-33% for Gallium body scan. The same author found that puncture biopsies are useful in 10-20% of cases and Kleijn26 found them so in 9-18%; while they found lymph node biopsy useful in 12 and 50% of the procedures performed. In our hospital, biopsies allowed diagnosis of 19.6% of the cases in the 1979-1987 study and 24% in the 2004-14 study, a figure that rises to 27.5 and 35.6%, respectively, if we consider patients undergoing laparotomy biopsy.

"Locator" tracking studies such as gammography and PET/CT have reported a diagnostic utility that is 25% for Gallium,27 38% for Indium 111,28 and up to 46-67% for PET/CT.22,29 In our hospital we do not have PET/CT, but we did tracking with Gallium 67 in 9 patients, which was decisive for diagnosis in only one patient and helpful as a locator in another two, for a diagnostic utility of 33.3%.

Recent studies report a significant decrease in the number of laparotomies performed, with 5% in the studies of Bleeker Rovers9 and Mete,27 with an unspecified number in the study of Vanderschueren8 (who reports the diagnostic utility of splenectomy in only 5 out of 192 cases studied). Mexican studies have shown a decrease from 50 to 10% of this surgery being done at the Instituto Nacional de la Nutrición from the 70’s to the 90’s;11,13 to a frequency higher than 30% at IMSS Centro Médico Nacional in the early 80’s.12 In the comparative analysis of the two studies from our hospital we observed a non-significant increase from 7.8 to 11% in the process being performed, with a diagnostic yield above 60% of cases in both groups, compared with 50% reported by other recent authors.9


FUO remains a challenge for clinicians and an entity that consumes the resources of health institutions. Knowledge of local epidemiology, a systematic approach, rational use of resources, and prudent behavior in non-serious cases, accompanied by regular clinical evaluation are the best tools for successful management of these cases.

The analysis of these two causal studies allow us to conclude that there has been a change in the clinical spectrum of this syndrome in our hospital with a rise in NIID, a decline in infections, and a steady rate of neoplastic diseases and undiagnosed cases, trends that are similar to some but not all recently published studies. The utility of laboratory and imaging tests is low, as literature reports, and it is necessary to perform locator tests that make possible invasive tests to confirm diagnosis in a significant proportion of cases. Studies are needed evaluating the diagnostic performance of new technologies such as PET/CT in Mexican patients with FUO. 

  1. Frati-Munari AC, Ariza-Andraca R, González-Gutiérrez T, Bañales-Ham MB, Chávez-Negrete A, Islas-Andrade S. Fiebre de origen oscuro. Análisis de 127 casos. Gac Med Mex 1988;124:426-34.
  2. Knockaert DC, Vanderschueren S, Blockmans D. Fever of unknown origin in adults: 40 yeras on. J Intern Med 2003;253:263-75.
  3. Petersdorf RG, Beeson PB. Fever of unexplained origin: Reporto n 100 cases. Medicine 1961;40:1-30.
  4. Durack DT, Street AC. Fever of unknown origin reexamined and redefined. Curr Clin Top Inf Dis 1991;11: 35-51.
  5. De Kleijn EMHA, Knockaert DC, Van der Meer JWM. Fever of unknown origin: A new definition and proposal for diagnostic work-up.
  6. Ergönül Ö, Willke A, Azap A, Tekeli E. Revised definition of fever of unknown origin: limitations and opportunities. Journal of Infection 2005;50:1-5.
  7. Petersdorf RG. Fever of unknown origin. An old friend revisited. Arch Intern Med 1992;152:21-2.
  8. Vanderschueren S, Knockaert DC, Adriaenssens T, Demey W, Durnez A, Blockmans D, et al. Arch Intern Med 2003;163:1033-41.
  9. Bleeker Rovers C, Vos FJ, De Kleijn EMHA, Mudde AH, Dofferhoff TSM, Richter C, et al. A prospective multicenter study on fever of unknown origin. The yield of a structured diagnostic protocol. Medicine 2007;86: 26-38.
  10. De Kleijn EMHA, Van der Meer JWM. Inquiry into the diagnostic workup of patients with fever of unknown origin. Netherlands Journal of Medicine 1997;50:69-74.
  11. De Lascuráin RM, Pérez Padilla JR. Fiebre de origen desconocido. Informe de 55 casos atendidos en el Instituto Nacional de la Nutrición de 1971 a 1977. Rev Invest Clin 1980;32:243-54.
  12. Moran S, Carcaño M, Halabe J, Lifshitz A. Más allá de la fiebre de origen oscuro. Un estudio de casos atendidos en el tercer nivel. Gac Med Mex 1992; 128:387-91.
  13. Molina-Gamboa J, Rivera-Morales I, Camacho Mezquita E, Ponce de León S. El espectro cambiante de la fiebre de origen oculto: Tendencias y comparación con series previas del Instituto Nacional de la Nutrición Salvador Zubirán. Rev Invest Clin 1994; 46:177-85.
  14. Larson EB, Featherstone HJ, Petersdorf RG. Fever of undetermined origin: diagnosis and follow-up of 105 cases, 1970-1980. Medicine 1982;61:269-92.
  15. Knockaert DC, Vanneste LJ, Vanneste SB, Bobbaers HJ. Fever of unknown origin in the 1980s. An update of the diagnostic spectrum. Arch Intern Med 1992;152: 51-55.
  16. De Kleijn EMHA, Vandenbroucke MV, Van der Meer JWM, and the Netherlands FUO study group. Fever of unknown origin (FUO) I. A prospective multicenter study of 167 patients with FUO, using fixed epidemiologic entry criteria. Medicine 1997;76:392-400.
  17. Efstathiou SP, Pefanis AV, Tsiakou AG, Skeva II, Tsioulos DIAchimastos AD, et al. Fever of unknown origin: Discrimination between infectious and non-infectious causes. Eur J Intern Med 2010;21:137-43.
  18. Colpan A, Onguru P, Erbay A, Akinci E, Aydin Cevik M, Sirri Eren S, et al. Fever of unknown origin: Analysis of 71 consecutive cases. Am J Med Sci 2007; 334:92-6.
  19. Baicus C, Bolosiu HD, Tanasescu C, Baicus A. Fever of unknown origin-predictors of outcome. A propsective multicenter study on 164 patients. Eur J Intern Med 2003;14:249-54.
  20. Liu KS, Sheng WH, Chen YC, Chang SC, Hsieh WC. Fever of unknown origin: a retrospective study of 78 adult patientes in Taiwan. J Microbiol Immunol Infect 2003;36:243-7.
  21. Naito T, Mizooka M, Mitsumoto F, Kanazawa K, Torikai K, Ohno S et al. Diagnostic work-up for fever of unknown origin: a multicenter collaborative retrospective study. BMJ open 2013;3:e003971.doi:10:1136/bmjopen-2013-003971.
  22. Hayakawa K, Ramasamy B, Pranatharthi H, Chandrasekar MD. Fever of unknown origin: An evidence based review. Am J Med Sci 2012;344:307-16.
  23. Arce-Salinas CA, Morales-Velázquez JL, Villaseñor-Ovies P, Muro-Cruz D. Clsssical fever of unknown origin (FUO): current causes in Mexico. Rev Invest Clin 2005;57:762-9.
  24. Tanveer M, Nabi-Dhobi G, Nabi-Koul A, Saleh T. Clinical profile of classical fever of unknown origin (FUO). Caspian J Intern Med 2014;5:35-9.
  25. De Kleijn EMHA, Van Lier HJJ, Van der Meer JWM, and the Netherlands FUO study group. Fever of unknown origin (FUO). II. Diagnostic procedures in a prospective multicenter study of 167 patients. Medicine 1997;76: 401-14.
  26. Mete E, Vanli E, Yemisen M, Balkan II, Dagtekinh H, Osaras R et al. The role of invasive and non-invasive procedures in diagnosing fever of unknown origin. In J Med Sci 2012;9:682-9.doi.10.7150/ijms.4591.
  27. De Kleijn EMHA, Oyen WJG, Claessens RAMJ, Corstens FHM, Van der Meer JWM. Utility of scintigraphic methods in patients with fever of unknown origin. Arch Intern Med 1995;155:1989-94.
  28. Blockmans D, Knockert D, Maes A, De Caestecker J, Stroobants S, Bobbaers H, Mortelmans L. Clinical value of 18F fluoro-deoxyglucose positrón emission tomography for patients with fever of unknown origin. Clin Infect Dis 2001;32:191-6.
  29. Balink H, Collins J, Bruyn G, Gemmel F. F-18 FDG/CT in the diagnosis of fever of unknown origin. Clin Nucl Med 2009;34:862-8

Conflict of interest statement: The authors have completed and submitted the form translated into Spanish for the declaration of potential conflicts of interest of the International Committee of Medical Journal Editors, and none were reported in relation to this article.

Enlaces refback

  • No hay ningún enlace refback.