Definitions. FUO indicates fever of unknown origin; asterisk, the study included only immunocompetent patients with a community-acquired illness; dagger, days comprise either in-hospital days or outpatient visits; and double dagger, the original definition of Petersdorf and Beeson1 required inpatient evaluation.
Composition of the cohort and of the subgroups according to the time of diagnosis. HIV indicates human immunodeficiency virus.
Vanderschueren S, Knockaert D, Adriaenssens T, Demey W, Durnez A, Blockmans D, Bobbaers H. From Prolonged Febrile Illness to Fever of Unknown OriginThe Challenge Continues. Arch Intern Med. 2003;163(9):1033-1041. doi:10.1001/archinte.163.9.1033
Epidemiological changes and the ongoing expansion of the diagnostic armamentarium warrant a regular update of the spectrum of diseases that present as prolonged febrile illnesses.
We prospectively collected a series of 290 immunocompetent patients referred to our university hospital between 1990 and 1999 with a febrile illness (temperature >38.3°C) of uncertain cause, lasting at least 3 weeks. Patients were categorized in 4 groups according to the timing of diagnosis: early diagnosis (within 3 in-hospital days or 3 outpatient visits), intermediate diagnosis (between days 4 and 7), late diagnosis (after day 7), and no diagnosis during index contact or follow-up.
A final diagnosis was established early in 67 patients (23.1%), intermediate in 38 (13.1%), and late in 87 (30.0%). In the remaining 98 (33.8%), no diagnosis was made. The cause of the fever remained obscure in 50 (47.6%) of 105 patients with episodic fever vs 48 (25.9%) of 185 patients with continuous fever (P<.001). Among the 192 patients with a final diagnosis, noninfectious inflammatory diseases represented the most prevalent diagnostic category (35.4%), surpassing infections (29.7%), miscellaneous causes (19.8%), and malignancies (15.1%). Fourteen disorders accounted for over 59% of diagnoses, whether diagnosis was reached early, intermediate, or late. Hematological malignancies made up 11.5% of diagnoses, but were responsible for 14 (58.3%) of the 24 fatalities related to the febrile illness. Of the 80 patients discharged alive without diagnosis and for whom follow-up was available, 3 died, but the deaths were considered to be unrelated to the feverish illness.
Prolonged febrile illnesses remain a diagnostic challenge. Despite the technological progress of the late 20th century, the origin of the fever remains elusive in many patients, especially in those with episodic fevers. Noninfectious inflammatory diseases emerge as the most prevalent diagnostic category.
IN 1961, Petersdorf and Beeson1 published their classic article on fever of unknown origin (FUO) and established criteria that have effectively delineated this entity: (1) an illness of at least 3 weeks' duration, (2) with fever (temperature >38.3°C on several occasions), and (3) no established diagnosis after 1 week of hospital investigation. This time-honored definition has enabled research on the topic and comparisons over time and between regions of the diseases and diagnostic categories that make up the FUO spectrum.2- 6 The first 2 criteria allow elimination of most acute, self-limited, frequently viral diseases and habitual hyperthermia, respectively. In 1991, Durack and Street7 suggested a modification of the third criterion. This revision was not grounded on clear scientific evidence, but responded to evolving trends in medical practice, including a shift toward outpatient management, advances in diagnostic techniques, and an accelerated pace of exploration. Accordingly, a prolonged febrile illness would classify as FUO if it fulfills the first 2 criteria and if the diagnosis remained uncertain despite appropriate investigations after at least 3 outpatient visits or at least 3 days in the hospital. This redefinition of FUO, however, has never been validated, and its impact on the case-mix is unknown. Moreover, given the increasing complexities of modern-day medicine, Durack and Street7 proposed to divide FUO into 4 groups: classic FUO, nosocomial FUO, neutropenic FUO, and human immunodeficiency virus (HIV)-associated FUO.
Herein, we present a prospectively collected series of 290 nonimmunocompromised patients presenting in the 1990s to a single tertiary care center with a community-acquired, prolonged, unexplained febrile illness fulfilling the first 2 prerequisites of the FUO definition and corresponding to the so-called classic FUO of Durack and Street.7 This cohort allows a comparison of patient groups according to the timing of diagnosis, as well as an update of the diagnostic spectrum of FUO.
From January 1990 through December 1999, we prospectively gathered a series of consecutive patients with prolonged unexplained fever at the time of referral. Patients either presented to the Unit of General Internal Medicine of the University Hospital Leuven (Leuven, Belgium) or were admitted to other units of internal medicine and came to our attention because our advice was sought. To be included in the final database, patients had to meet the first 2 criteria of the FUO definition of Petersdorf and Beeson1: (1) duration of illness of more than 3 weeks before diagnosis and (2) repeatedly documented body temperature exceeding 38.3°C.1 Patients with nosocomial fever or with known HIV infection or immunocompromise were excluded. Immunocompromise encompassed the following conditions: neutropenia (white blood cell count <1.0 × 103/µL and/or granulocyte count <0.5 × 103/µL); hypogammaglobulinemia (IgG <50%); or the intake of the equivalent of more than 10-mg prednisone or other immunosuppressive agents for at least 2 weeks. A minimal diagnostic workup, including history review, clinical examination, routine laboratory tests, urinalysis, cultures, serology, chest radiography, and abdominal ultrasonography, was required. General guidelines for the investigation of FUO were available,8 but no rigid protocol was followed. From the beginning of 2000 onward, when the log was complete, the patient records were retrieved and evaluated for inclusion in the final database. Demographic data, details of history, and findings from physical examination and laboratory and technical investigations were registered in a structured data collection form. Episodic fever was defined as at least 2 episodes of fever, with fever-free intervals of at least 2 weeks and seeming remission of the underlying illness.9
The final diagnosis established at discharge or during follow-up constituted the main outcome measurement. Only diagnoses that were sufficiently validated and refined were retained. The test that most convincingly ascertained the final diagnosis was determined. According to the timing of diagnosis, patients were categorized in 4 groups: early diagnosis (within 3 days of patient evaluation, either as inpatient or outpatient); intermediate diagnosis (between days 4 and 7); late diagnosis (after day 7); or no diagnosis during index contact or follow-up. When a diagnosis was strongly suspected and subsequently confirmed by guided diagnostic action, the time of strong suspicion was considered the time of diagnosis. The patients with a late diagnosis or without diagnosis thus meet all criteria of the FUO definition of Petersdorf and Beeson,1 with the exception that outpatient evaluation was allowed (Figure 1). Petersdorf endorsed this slight digression from his original definition.10 The patients with an intermediate or late diagnosis or without a diagnosis comply with the revised definition of classical FUO of Durack and Street.7 The vital outcome was established by examining the patient files or computerized records, focusing on the patients with follow-up in our university hospital. For patients dying after discharge, it was determined whether the febrile illness caused the death.
The causes of FUO were classified in 4 diagnostic categories: (1) infections; (2) malignancies; (3) noninfectious inflammatory diseases, a category coined by De Kleijn et al11 that comprises connective tissue diseases, vasculitides, and granulomatous disorders; and (4) miscellaneous causes, including drug-related fever, habitual hyperthermia, and factitious fever. Historical series encompassing more than 100 patients satisfying the FUO definition of Petersdorf and Beeson1 and presenting to university hospitals1,11- 15 were reclassified according to the previously mentioned diagnostic categories and used for comparison. Descriptive terms such as "granulomatous hepatitis" and "nonspecific myositis" were not considered final diagnoses and were reclassified into the "no diagnosis" category. Because the present study was merely observational and did not demand deviations from standard care, no informed consent was sought.
We used the Pearson χ2 test or the Fisher exact test to compare categorical variables. Continuous variables are presented as medians and interquartile ranges and were compared with the use of 1-way analysis of variance or the Kruskal-Wallis test and, in pairwise analyses, with the use of the Mann-Whitney test. We applied logistic regression to select variables that predict whether a final diagnosis would be made and whether a long-lasting fever would go on to meet the FUO definition of Petersdorf and Beeson.1 Statistical analyses were performed using SPSS software version 8.0 (SPSS Inc, Chicago, Ill). All statistical testing was performed using 2-tailed tests, with significance at P<.05.
The prospectively collected register listed 483 patients; 193 were excluded. Reasons for exclusion were inadequate documentation of temperature surpassing 38.3°C on at least 3 occasions (67 patients); illness resolving within 3 weeks without recurrence or diagnosis established within 3 weeks of symptom onset (61 patients); insufficient basic workup (eg, only 1 outpatient visit) (32 patients); diagnosis already known at referral (8 patients); nosocomial fever (13 patients); index contact in 1989 (6 patients); and known immune deficiency, HIV seropositivity, or neutropenia (6 patients). Of the 483 patients, 290 matched all study criteria and were retained for analysis (Figure 2). During the index contact, the cause of the prolonged febrile illness was elucidated within 3 days in 67 patients (early diagnosis in 23.1%), between days 4 and 7 in 38 (intermediate diagnosis in 13.1%), and after day 7 in 68 (late diagnosis in 23.4%). In 19 of the 117 patients discharged without diagnosis, a final diagnosis could be established during follow-up. These patients were added to the "late diagnosis" category, which thus totaled 87 patients (30.0%). The discharge diagnosis of pneumonia in 1 patient was subsequently changed into pulmonary tuberculosis as culture results became available. A final diagnosis was absent in 98 patients (33.8%).
Table 1 gives baseline variables and the duration of hospitalization of the group of 290 patients and of the 4 subgroups (with early, intermediate, late, and no final diagnosis). Median age was comparable in the 4 subgroups. Female patients predominated in the groups with early and intermediate diagnosis, while the reverse was true for the groups with late or without diagnosis. General practitioners referred most patients in the groups with early and intermediate diagnosis, while most patients in the groups with late diagnosis and without diagnosis were specialist referrals. Whereas more than a fifth of the group with early diagnosis was managed exclusively through ambulatory care, this became an exception as the time to diagnosis increased. Not surprisingly, the length of hospital stay was longest in patients with late diagnosis. The duration of illness before presentation, baseline laboratory inflammatory parameters, and hemoglobin level did not differ significantly between the 4 groups, although C-reactive protein tended to be the highest and hemoglobin level the lowest in the group with a late diagnosis.
Table 2 lists the final diagnoses and the diagnostic categories in the 192 patients with diagnosis, subdivided according to the timing of diagnosis. Overall and in the group with late diagnosis, noninfectious inflammatory diseases constituted the most prevalent diagnostic category. In the group with early diagnosis, infection was the most common diagnosis. Neoplasms tended to be diagnosed late, while the share of miscellaneous disorders remained fairly constant. Considering individual disorders, some entities (eg, endocarditis) tended to be detected early, whereas others (such as tuberculosis and sarcoidosis) escaped immediate diagnosis. Fourteen entities (endocarditis, tuberculosis, abdominal abscess, Epstein-Barr and cytomegalovirus infection, lymphoma, leukemia, adult-onset Still disease, systemic lupus erythematosus, polymyalgia rheumatica and giant cell arteritis, sarcoidosis, Crohn disease, subacute [De Quervain] thyroiditis, habitual hyperthermia, and drug fever) accounted for at least 4 cases each. Together, these disorders made up 127 (66.1%) of the 192 diagnoses; 48 (71.6%) of the 67 early diagnoses, 27 (71.0%) of the 38 intermediate diagnoses, and 52 (59.8%) of the 87 late diagnoses.
Table 3 gives the decisive method of diagnosis in the 192 patients in whom a diagnosis was obtained. Biopsy was the most rewarding technique, especially in the groups with intermediate or late diagnosis. Biopsy of lymph node (n = 10), temporal artery (n = 9), and bone marrow (n = 8) in particular provided a high diagnostic yield. In 5 cases autopsy revealed the diagnosis. The clinical course led to the diagnosis in 22.9% of patients. Microbiological analyses (cultures and serology combined) also had a reasonable diagnostic yield (13.6%). Echocardiography and immunological techniques (including antinuclear antibody testing) proved useful during the early phase of investigation. Of course, imaging techniques (including radiological and scintigraphic methods), although infrequently leading in isolation to a definite diagnosis, were often contributory.
The vital outcome of the patients is depicted in Table 4. Sixteen patients died during the index admission. In 13, the cause of death was known: non-Hodgkin lymphoma and acute leukemia (4 cases each); disseminated tuberculosis (2 cases); pulmonary embolism, Salmonella aortic root prosthesis infection, and Campylobacter species prosthetic aortic valve endocarditis (1 case each). In 5 of these patients, autopsy revealed the final diagnosis (pulmonary embolism in 1 case and miliary tuberculosis and acute aleukemic leukemia in 2 cases each). In 3 cases, no diagnosis could be achieved despite autopsy. Follow-up data after discharge were available for 233 patients (85.0%). Median (interquartile range) duration of follow-up was 810 (180-2085) days in the total group and 574 (125-2052) days in the group without final diagnosis. Eighteen patients died during the follow-up period. In 11 of them, the cause of death was related to the febrile illness. The diagnosis in these cases included non-Hodgkin lymphoma (3 cases); leukemia (2 cases); and angioimmunoblastic lymphadenopathy, acquired immunodeficiency syndrome, disseminated adenocarcinoma with unknown primary tumor, rectal carcinoma, Wegener granulomatosis, and nonclassified vasculitis (1 case each). Hence, hematological malignancies, while accounting for 11.5% of diagnoses, were responsible for 14 (58.3%) of the 24 fatalities related to the febrile illness.
In 80 of the 95 patients without definite diagnosis, follow-up information after discharge was available. Death occurred in 3, but was presumably not linked to the febrile illness. One patient died of meningococcal meningitis 1 year after the index admission, another of bronchogenic carcinoma 20 months later, and a third of non-Hodgkin lymphoma 6 years later. Of the other 77 patients, 63 were alive and well and 14 were alive with recurrent symptoms. In 7 patients of the group without definite diagnosis, severe illness or clinical deterioration during the index admission necessitated the initiation of corticosteroid therapy.
Table 5 compares the cohorts of patients that meet the criteria for FUO of Petersdorf and Beeson1 and of Durack and Street.7 While the former only consists of the groups with late diagnosis and without diagnosis, the group with intermediate diagnosis is added to the latter ( Figure 1 ). No significant difference in proportion of diagnostic categories among diagnosed cases was found (P = .46). Table 2 reveals that all common diagnoses uncovered between days 3 and 7 also rank among the diseases diagnosed later on.
In univariate analysis, the groups with and without diagnosis did not differ significantly as to age, duration of illness before presentation, baseline erythrocyte sedimentation rate, and hemoglobin and C-reactive protein levels. A diagnosis was reached in 99 (60.4%) of the 164 male patients vs 93 (73.8%) of the 126 female patients (P = .02). A cause could be established in 55 (52.4%) of the 105 patients with episodic fever vs 137 (74.1%) of the 185 patients with continuous fever (P<.001). A diagnosis was revealed in 106 (71.6%) of the 148 patients referred by a general practitioner vs 86 (60.6%) of the 142 referred by a specialist (P = .048). When these 8 parameters were entered in a logistic regression analysis to select variables that might predict whether a final diagnosis would be made, only the periodicity of the fever emerged as statistically significant (odds ratio, 0.40; P = .01).
Logistic regression with the same independent variables was also applied to predict which long-lasting fevers would evade diagnosis during 1 week of exploration and thus progress to FUO as defined by Petersdorf and Beeson.1 Significant predictors were recurrent fever compared with continuous fever (odds ratio, 4.0; P = .001) and male vs female sex (odds ratio, 2.1; P = .04).
Table 6 compares the diagnostic categories of the present study with those reported from other university centers including at least 100 patients and applying the FUO criteria of Petersdorf and Beeson.1,11- 15 A high proportion of "no diagnosis" is noted in the present study (53.0%). Among the diagnosed cases, the "noninfectious inflammatory diseases" category has gradually replaced the infectious category as the largest.
Advances in serological, immunological, imaging, and genetic techniques continuously change the face of diagnostic medicine. These technological evolutions spark the hope that FUO might become a thing of the past. The present series of 290 immunocompetent patients, who presented during the 1990s to a single university hospital with a prolonged community-acquired unexplained febrile illness, contradicts this assumption. The cause of the illness remained obscure in 33.8% of the entire cohort and in 53.0% of the patients fulfilling the classic FUO criteria of Petersdorf and Beeson.1
This observation is not easily explained. Our center's long-term experience with and commitment to FUO research9,14,16- 18 and the diversity of diseases, including rare disorders, that were uncovered argues against lack of diagnostic acumen. In fact, the present series introduces into the FUO arena new disease entities (linear IgA dermatosis19), new disease presentations (Still disease presenting as neutrophilic meningitis20), and new diagnostic techniques (fluorodeoxyglucose positron emission tomography21).
The number lost to follow-up in the no diagnosis category was too small to substantially affect the proportion with persistently unknown origin of fever, even in the hypothesis that a final diagnosis would have emerged in a substantial amount of these patients. In addition, the median duration of follow-up was long enough to contradict this hypothesis.
Another explanation refers to the diagnostic strictness we applied. A few examples may strengthen this point. Entities such as granulomatous hepatitis, inflammatory jejunitis, and macrophage activation syndrome were considered to be insufficiently etiologically refined and entered in the no diagnosis category. Feverish illnesses that waned spontaneously during the exploratory phase were not classified as benign infections in the absence of decisive microbiological evidence. A urinary tract infection was not regarded as the final diagnosis when antibiotic therapy was associated with resolution of the fever, unless parenchymatous involvement was documented. If defervescence coincided with the discontinuation of a drug therapy but rechallenge was uncomplicated, drug fever was not withheld as a diagnosis. We only accepted diagnoses that lack persuasive confirmatory tests (eg, adult-onset Still disease, Behçet syndrome, and polymyalgia rheumatica) if sufficient standard clinical criteria were met and follow-up allowed excluding other conditions.
Referral pattern might also influence the number of undiagnosed cases. Indeed, nearly half of the patients in our study were specialist referrals. Referral by specialists was associated with a decreased likelihood of reaching a diagnosis in univariate but not in multivariate analysis. The one factor that emerged from the logistic regression analysis as a predictor of diagnostic failure was the periodicity of the fever. In one quarter of patients with continuous fever, but in almost a half of patients with episodic fever, the cause remained uncertain. This phenomenon was also observed in other recent FUO series.11,14 In the series gathered in the 1980s in our center, no diagnosis was found in 51% of patients with recurrent fever and in 18% of patients with continuous fever.14 De Kleijn et al11 failed to establish a diagnosis in 50% of patients with recurrent fever vs 20% of patients with continuous fever. The proportion of episodic fevers in these previous FUO populations (22.6% and 33.5%, respectively) was lower than in ours (42.2%). Thus, this high percentage of episodic fevers may have lowered the diagnostic yield in the present cohort.
Other authors have attributed the high proportion of undiagnosed cases in their surveys to the early introduction of advanced diagnostic techniques, which may yield a diagnosis before the classic FUO criteria, including the requirement for 1 week of investigation, are met.11 However, if we consider our entire study population with prolonged fever, including the subgroups with a diagnosis within the first week, the percentage of undiagnosed cases (33.8%) remains high. In the same line of reasoning, one may argue that nowadays patients tend to seek medical advice early and that a diagnosis is frequently established before 3 weeks have elapsed. Only the difficult, hard-to-diagnose cases would remain. Some authors have advocated that a 2-week history of fever without apparent cause may be adequate for FUO.22 However, this proposed redefiniton has not been validated, and we do not endorse it because, to paraphrase Petersdorf,10 a hodgepodge of diagnoses risks to creep into FUO series when more short-lived febrile illnesses are included. Consequently, reducing the required duration of illness from 3 to 2 weeks may inflate rather than deflate the number without clear-cut diagnosis.
Another possible reason of the seemingly high diagnostic failure rate relates to the observation that most patients with FUO who lack a diagnosis fare well.12,17 This knowledge may inspire diagnostic abstinence in clinically stable patients once diagnoses with immediate therapeutic or prognostic implications have been ruled out to a reasonable extent. The finding that the median duration of hospitalization in our patients with FUO (17 days) was shorter than that in other recent series (25 and 27 days)11,14 supports this hypothesis. This argument holds especially for patients with episodic fever who, by definition,9 are asymptomatic in between febrile episodes and who are rarely inclined for another diagnostic round once the symptoms have subsided. In our series also, the vital outcome of the patients who left the hospital without diagnosis was favorable. Spontaneous resolution occurred in most. Thus, time more often than not proved to be the ally of the physician. Of course, the same expectant attitude cannot be extrapolated to patients with clinical deterioration or to other febrile populations, such as patients with HIV, neutropenia, or nosocomial or acute fever, in whom a continued arduous attempt to find and treat the cause of the fever is often vitally important. These latter categories were excluded from analysis in our final database.
While some of the above-mentioned explanations may to a certain extent account for the large fraction of persistently unexplained fevers, part of it is genuine: some prolonged fevers remain enigmatic despite vigorous diagnostic effort. The 8 patients who died in the hospital before a unifying diagnosis exemplify this: autopsy revealed the cause in only 5. An overview over the last 5 decades of the large FUO series from university-based centers in western countries shows a clear trend toward an increasing amount of unresolved cases1,11- 15 (Table 6). Thus, some fevers remain of unknown origin and represent a source for humility on the part of the diagnostician, but may at the same time serve as an impetus for continued research. Apparently, in a subset of patients, an as yet unidentified noxious, possibly infectious or immunological, stimulus against a vulnerable background may incite an exaggerated, prolonged but ultimately transitory, inflammatory response.
Predicting which patients with protracted fever will defy diagnosis is difficult. Some studies have found that an elevated sedimentation rate or abnormal hemoglobin level increased the likelihood of reaching a diagnosis.23 We could not confirm this association. Laboratory inflammatory parameters at presentation in our patients without diagnosis were not significantly different from those with diagnosis. Thus, apart from the periodicity of the fever, few baseline parameters enable the prediction of when prolonged fevers become FUO.
Durack and Street7 quite reasonably have proposed to consider nosocomial, neutropenic, and HIV-associated FUO as distinct groups different from classic FUO. In addition, they suggested to reduce the exploratory phase from 1 week to 3 days before a prolonged unexplained fever would qualify as FUO.7 This redefinition of classic FUO has not been validated before. Per se, it will reduce the proportion of unresolved cases (from 53.0% to 43.9% in the present series) (Table 5). Our study shows that the revised definition does not alter the relative contribution of the different diagnostic categories in the cases with a final diagnosis. Moreover, all prevalent diagnoses uncovered between days 3 and 7 also rank among the diseases diagnosed later on. Thus, the updating of the definition does not seem to affect the case-mix nor the diagnostic approach. Nonetheless, we argue that any quantitative criterion based on a certain number of days to complete a workup will remain arbitrary and subject to local logistics and expertise. Instead, we suggest adherence to a qualitative criterion that specifies which examinations are necessary before an unsolved prolonged fever deserves the label FUO. This proposition will also circumvent the discussion whether a day of hospital evaluation and an outpatient visit should be gauged equally. Standard initial diagnostic protocols have been proposed.4,11,24 Each protocol should be adapted according to regional infectious epidemiological factors. For example, leishmaniasis and Q fever are common causes of FUO in Spain,13 but are extremely rare in more northern European countries and the United States.
Comparison of our present FUO series with historical surveys from university hospitals including at least 100 patients1,11- 15 reveals, apart from the rise in undiagnosed cases, an increase in the noninfectious inflammatory disease category (Table 6). This category now surpasses infectious diseases as the most important class. Interestingly and not unexpectedly, infections remain the prevailing category in our early diagnosis group, as in earlier FUO studies.1,12,13 Other authors who have presented an update of their previous FUO series also noted an increase in noninfectious inflammatory diseases.15,25 In community hospitals, infections may remain the most common category.26
We have adopted the neutral term noninfectious inflammatory diseases, as suggested by De Kleijn et al11 in lieu of rheumatic diseases, multisystem diseases, dyscollagenosis, collagen diseases, collagen vascular diseases, connective tissue diseases, and autoimmune diseases. This category also comprises granulomatous disorders such as sarcoidosis and inflammatory bowel disease that used to be listed under the miscellaneous disorders. Nonetheless, no terminology is perfect, and some disorders of the miscellaneous category (eg, subacute thyroiditis and primary sclerosing cholangitis) are also noninfectious and inflammatory, albeit with a stricter target.
Interestingly, although reported causes of FUO exceed 200,4 a limited list of disorders accounted for the majority of our diagnoses. These include endocarditis, tuberculosis, abdominal abscess, Epstein-Barr and cytomegalovirus infection, lymphoma, leukemia, adult-onset Still disease, systemic lupus erythematosus, polymyalgia rheumatica and giant cell arteritis, sarcoidosis, Crohn disease, subacute (De Quervain) thyroiditis, habitual hyperthermia, and drug fever. Together, these diagnoses were well represented in the early, intermediate, and late diagnosis groups. Petersdorf and Beeson1 have already emphasized that most patients with FUO are not experiencing unusual diseases, but rather exhibit atypical manifestations of common illnesses. From a prognostic point of view, hematological malignancies, especially non-Hodgkin lymphoma, were the most dreaded diagnoses. Constituting 11.5% of diagnoses, hematological malignancies caused more than half of all deaths related to the febrile illness. This observation agrees with the reported increase of incidence and death rate of non-Hodgkin lymphoma that have nearly doubled since 1970.27
The strength of this study is the inclusion of immunocompetent patients with prolonged unexplained febrile illness irrespective of a predetermined timing of diagnosis. It shows that certain entities such as bacterial endocarditis, systemic lupus erythematosus, and infectious mononucleosis can be and should be detected early, before the classic FUO criteria are fulfilled. On the basis of the present cohort, we cannot estimate the true incidence of prolonged fevers of uncertain cause, since almost half of the patients were secondary referrals, and, on the other hand, undoubtedly, not all cases admitted to our university hospital came to our attention.
Thus, at the beginning of the 21st century, protracted fever of obscure cause remains one of the more daunting challenges facing the physician. Continuous shifts in prevailing diseases and diagnostic categories warrant a regular update of the spectrum as knowledge of the array of causes forms the basis of a rational approach. The increasing sophistication of diagnostic technology has as yet failed to improve the diagnostic acumen. Some prolonged fevers remain enigmatic, but this should not invoke dismay but rather inspire ongoing research to unravel underlying pathophysiological mechanisms.
Corresponding author and reprints: Steven Vanderschueren, MD, PhD, General Internal Medicine, University Hospital Leuven, Herestraat 49, B-3000 Leuven, Belgium (e-mail: Steven.Vanderschueren@uz.kuleuven.ac.be).
Accepted for publication July 30, 2002.
We are indebted to Willy Peetermans, MD, PhD, and Eric Van Wijngaerden, MD, PhD, who were involved in the clinical care of the patients, to Alexander Wilmer, MD, PhD, for review of the manuscript, and to the study nurses Helga Ceunen and Marina Lejeune for data input in a computerized database.