Adult-onset Still disease (AOSD) is a systemic inflammatory condition of unknown etiology. Since its first description in 1971, the characteristics of approximately 300 cases have been reported worldwide1,10. Clinically it resembles pediatric Still disease. Patients typically present with high spiking fever, arthralgia or arthritis, transient maculopapular rash, lymphadenopathy, hepatosplenomegaly, serositis, and pharyngitis. Marked neutrophilic leukocytosis and elevated erythrocyte sedimentation rate are generally present, as well as mild elevations in hepatic enzymes. Rheumatoid factor and antinuclear antibodies are characteristically absent11,13,15.
AOSD is not frequent: its prevalence has been estimated to be lower than 1 case per 100,000 persons7,16. This fact, in addition to the ambiguity of its definition and the lack of serologic markers, makes diagnosis difficult. Early disease, especially when patients present as fever of unknown origin (FUO), can be easily mistaken for other conditions such as malignant lymphoma or chronic infections. Thus AOSD is usually considered an exclusion diagnosis. These matters are reflected in the fact that 7 sets of criteria for AOSD have been proposed2-6,14,17. In 1996, Masson et al published a study comparing the 6 existing criteria8. The Yamaguchi criteria17 were the most sensitive. Nevertheless, a new set of classification criteria that considers glycosylated ferritin as a major criterion was proposed by Fautrel et al4. According to their work, their criteria are the most accurate: their sensitivity was 80% and their specificity 98.5%4. In spite of that, in clinical practice AOSD is considered an exclusion diagnosis, particularly when it presents as FUO.
FUO is a complex syndrome with an extensive list of differential diagnoses, in which up to 20% of cases lack an etiologic diagnosis after a long workup9,12. Furthermore, patients with FUO are usually subject to expensive and invasive diagnostic procedures in order to exclude infection and malignancy. In these cases, is AOSD an exclusion diagnosis? If so, which diseases must be excluded before the diagnosis is established? May a patient with FUO who fulfills AOSD classification criteria be diagnosed as such before excluding other diseases? We conducted the current study to answer those questions.
PATIENTS AND METHODS
We conducted a retrospective, case-control study. We reviewed the clinical records of patients diagnosed with AOSD between 1991 and 2001 in our institution. Patients hospitalized for FUO served as controls. To identify them, we reviewed the records of patients admitted for fever during the same period. Of 2104 patients, we randomly chose 1000; 156 of them fulfilled criteria for FOU (see below). We excluded neutropenic and human immunodeficiency virus (HIV)-infected patients, as well as patients with an established diagnosis that could explain the febrile syndrome. The remaining 135 were included in the database as controls (about 5 per case). In our institution, during the decade considered, the delay to diagnosis in patients undergoing study for FUO was approximately 2 weeks. Thus, FUO was defined according to Petersdorf and Beeson12. Briefly, FUO was diagnosed when a patient was without an established diagnosis after 1 week of hospital study, with temperature higher than 101 °F (38.3 °C) for at least 3 weeks.
We considered the diagnosis established by the attending physicians. Infectious diseases were diagnosed by means of positive cultures and/or serology; malignant conditions were diagnosed by means of biopsies; rheumatic and autoimmune diseases by rheumatologists. All patients with rheumatic diseases fulfilled the American College of Rheumatology classification criteria (when available). AOSD was diagnosed by consensus (of expert rheumatologists) in patients with compatible clinical and laboratory findings, after the exclusion of infectious and malignant conditions. None of the published criteria was used to diagnose AOSD. In the cases in which no consensus was reached, patients remained without a diagnosis and their clinical course was followed until enough clinical and laboratory data permitted a diagnosis. Only patients in whom a definitive diagnosis was established were considered for the study.
Clinical characteristics and laboratory parameters were recorded for each case and control. We considered the clinical manifestations present from the beginning of the disease until its diagnosis. The highest recorded temperature was considered. Weight loss was defined as more than 10% of body weight loss since the beginning of the disease. Laboratory findings and physical examination at the time of hospital admission were documented.
Patients were classified into 5 groups: 1. Infectious diseases; 2. Malignant conditions; 3. Autoimmune diseases; 4. No diagnosis; 5. AOSD. Clinical and laboratory data from the different groups were compared. A second analysis considered only the patients with the most common diseases (tuberculosis, Hodgkin lymphoma, non-Hodgkin lymphoma, systemic lupus erythematosus, and AOSD).
Differences in baseline characteristics between groups of patients were evaluated by analysis of variance (ANOVA). Odds ratios (OR) with 95% confidence intervals (95% CI) were calculated by multiple logistic regression analyses to estimate the relative risk of each variable for patients with AOSD, compared with patients with other diagnoses. Cutoff points for leukocytosis and neutrophilia, as well as the threshold of the clinical scale, were calculated using ROC curves. We used SPSS version 10.1.0 (SPSS, Chicago, IL) for all analyses.
The study included 161 cases of FUO, 26 of which were cases with AOSD. Among the controls, 30 patients (22.2%) were in Group 1 (infectious diseases), 39 (28.8%) in Group 2 (malignant conditions), and 43 (31.8%) in Group 3 (autoimmune diseases). In 23 patients (17.0%) the diagnostic workup yielded no diagnosis (Group 4).
Group 1, infectious diseases, included 4 patients with abscesses (2 intraabdominal, 1 hepatic, 1 renal), 6 patients with bacteremia (1 Enterococcus species, 1 Candida species, 2 Salmonella typhi, 2 Salmonella species), 4 patients with brucellosis, 3 patients with endocarditis, and 13 patients with tuberculosis (2 pulmonary, 11 extrapulmonary).
Group 2, malignant conditions, comprised 2 patients with solid tumors (gastric and renal adenocarcinomas), 18 patients with Hodgkin lymphoma, 18 patients with non-Hodgkin lymphoma, and 1 patient with acute leukemia.
Group 3, autoimmune diseases, included 1 patient with rheumatoid arthritis, 1 with primary biliary cirrhosis, 2 with mixed connective tissue disease, 2 with spondyloarthropathy, 1 with adjuvant-induced disease, 2 with familial Mediterranean fever, 2 with autoimmune hepatitis, 1 with relapsing polychondritis, 4 with polymyalgia rheumatica, 1 with Sjögren syndrome, 1 with sarcoidosis, and 25 with systemic lupus erythematosus (including 5 patients with late-onset lupus).
General and Demographic Data
The mean age of patients from each group is presented in Table 1. Patients with AOSD were younger than patients in Groups 1, 2, and 3 (28.5 ± 12.8 yr versus 42.9 ± 19.1 yr, 42.1 ± 18.5 yr, and 40.8 ± 17.4 yr, respectively; p < 0.01). Age did not differ significantly between patients with AOSD and patients from Group 4 (no diagnosis). The percentage of female patients in Group 3 (autoimmune diseases) was higher than in the other groups. There were no statistically significant differences in the duration of the disease (see Table 1). Patients from all groups required extended hospitalizations; the length was longer in the group of patients with AOSD (23.7 ± 42 d) than in the other groups. However, the difference was not statistically significant. Follow-up time (defined as months between the formulation of the diagnosis and the last time the patient was seen in the institution) for patients with AOSD was 39.6 ± 30.6 months.
Clinical Characteristics Among Different Diagnostic Categories
Arthralgia and arthritis were more common in patients with AOSD and with autoimmune diseases than in patients from other groups (p < 0.05). It is noteworthy that 70% of patients from the group with no diagnosis (Group 4) complained of arthralgia. High fever was common to all groups. As expected, pharyngitis and typical Still rash were significantly more common in patients with AOSD than in other groups (p < 0.01). Peculiarly, myalgia was more frequently observed in patients with AOSD than in patients with infectious, malignant, and autoimmune conditions (p < 0.05, p < 0.01, p < 0.05, respectively). Weight loss was common in patients from all groups, disregarding the underlying condition. It was more frequent, however, in patients with malignant diseases than in patients with AOSD (85% versus 58%, p < 0.05). Lymphadenopathy, although predominantly associated with malignant conditions, was recorded in approximately half the patients from all groups. Splenomegaly was more common in patients with malignant diseases and AOSD than in patients with other conditions. On the other hand, hepatomegaly was a rather nonspecific finding. Pleuritis and pericarditis behaved in an unexpected manner in patients with AOSD. Although they were not frequently observed, they were always present synchronously. This information is presented in Table 2.
Laboratory Findings Among Different Diagnostic Categories
Low-grade anemia was a common finding in all patients, and there were no statistically significant differences in hemoglobin levels between groups. Leukocyte counts did not differ between the patients of the first 4 groups (8.8 ± 4.7, 7.8 ± 5.0, 7.2 ± 4.9, 11.0 ± 5.6, respectively), but were significantly higher in patients with AOSD (16.5 ± 5.6, p < 0.001). We defined leukocytosis as a leukocyte count higher than 10,500 cells per μL. It was present in 85% of patients with AOSD, 27% of patients with infectious diseases, 22% of patients with malignant conditions, and 19% of patients with autoimmune diseases (p < 0.001). Neutrophilia was defined as a neutrophil count higher than 9500 cells per μL. It was present in 85% of patients with AOSD. The difference was statistically significant when compared with the patients of the other disease groups (p < 0.001). Eosinophil count was normal in all patients (data not shown). Lymphopenia was a frequent finding in all groups and did not associate with any particular condition. The mean platelet count was higher in AOSD than in other conditions (p < 0.05); however, when defined as a platelet count higher than 450,000 per μL, thrombocytosis was as common in AOSD as in autoimmune and infectious diseases. On the other hand, thrombocytopenia was not found in a single patient with AOSD. Low-level elevation in liver enzymes was a constant finding in all groups. Hypoalbuminemia was a frequent finding in all groups; nevertheless, it was detected in all the patients with AOSD. The difference was statistically significant (p < 0.05). The mean globulin level was higher in patients with AOSD than in patients with malignant and autoimmune diseases (p < 0.05). Accordingly, hyperglobulinemia was more frequently found in patients with AOSD than in patients from other groups. However, the difference was not statistically significant. LDH elevation was found in all groups. This information is presented in Table 3.
Antinuclear antibodies (ANA) were positive in 20/39 patients with autoimmune diseases, and in 3/23 patients with AOSD. All 3 exhibited low-titer (≤1:160) speckled pattern ANA. Anti-dsDNA antibodies (ELISA) were positive in 4/17 patients with AOSD; rheumatoid factor was positive in 5/23 patients with AOSD. The titers of anti-dsDNA antibodies and rheumatoid factor were low in all patients with AOSD. Only 1 patient (of 17) with AOSD had hypocomplementemia (due to low C4).
Clinical and Laboratory Findings in Patients With Defined Diseases
To evaluate homogeneous groups, we selected the patients with the most prevalent diseases. We thus compared the clinical characteristics and laboratory findings from 13 patients with tuberculosis, 18 patients with non-Hodgkin lymphoma, 18 patients with Hodgkin lymphoma, and 25 patients with systemic lupus erythematosus, with the patients with AOSD.
When compared to patients with tuberculosis, non-Hodgkin lymphoma, and Hodgkin lymphoma, arthralgia, arthritis, pharyngitis, and Still rash were significantly more common in patients with AOSD (p < 0.05, p < 0.001, p < 0.001, p < 0.01, respectively). Likewise, pharyngitis, Still rash, and splenomegaly were more frequent in patients with AOSD than in patients with systemic lupus erythematosus (p < 0.05, p < 0.001, p < 0.05, respectively).
As expected, leukocytosis and neutrophilia were significantly more common in patients with AOSD than in patients with tuberculosis, non-Hodgkin lymphoma, Hodgkin lymphoma, and systemic lupus erythematosus (p < 0.001). Thrombocytosis and hyperglobulinemia were more frequent in patients with tuberculosis and AOSD than in patients with other diseases (Table 4). There were no significant differences in other clinical or laboratory parameters between groups (data not shown).
Univariate and Multivariate Analysis
We compared clinical and laboratory variables from patients with AOSD with those of patients from the other groups in a univariate analysis. The variables that displayed a stronger association with the presence of AOSD were neutrophilia (OR, 28; 95% CI, 9-90), Still rash (OR, 18; 95% CI, 6-52), arthritis (OR, 18; 95% CI, 5-62), leukocytosis (OR, 15; 95% CI, 5-48), pharyngitis (OR, 10; 95% CI, 4-26), and arthralgia (OR, 8, 95% CI, 2-36). Table 5 presents the results of the multivariate analysis. Neutrophilia (OR, 18; 95% CI, 3-93), arthritis (OR, 8; 95% CI, 1-49), pharyngitis (OR, 7; 95% CI, 1-30), and splenomegaly (OR, 5; 95% CI, 1-26) remained strongly associated with the presence of AOSD.
Clinical Scale for the Diagnosis of AOSD
Using the results of the multivariate analysis, we designed a clinical scale that can be applied to patients with FUO to identify the subset of patients that can be diagnosed as having AOSD without further workup. The scale comprises 5 criteria (arthritis, pharyngitis, Still rash, splenomegaly, and neutrophilia). Four of them were identified in the multivariate analysis. Even though the association strength between Still rash and AOSD was borderline, we decided to include it in the clinical scale because we believe that due to the retrospective nature of the study, the prevalence of Still rash might be underrepresented in the database. We believe that if specifically sought, and adequately defined, its diagnostic accuracy might be higher.
The presence of each of the criteria confers a different number of points; the points are added to obtain a total score (Table 6). If the score equals ≥30, AOSD can be diagnosed.
Performance of the Clinical Scale and the Yamaguchi Criteria in Cases and Controls
We applied the clinical scale and the criteria of Yamaguchi and colleagues17 to cases and controls. In this analysis, we did not consider the patients from Group 4 (no diagnosis). Likewise, 12 patients were not included because of incomplete data. Thus, the criteria and the clinical scale were applied to 126 records. The Yamaguchi criteria had a sensitivity of 82.6% and specificity of 91%, with a positive predictive value of 70.3% and negative predictive value of 95.2%; the positive likelihood ratio for the Yamaguchi criteria was 10.3 (4.7-17.1). The clinical scale had a sensitivity of 76.9% and specificity of 98%, with a positive predictive value of 90.9% and negative predictive value of 94.2%. Its positive likelihood ratio was 38.5 (9.5-152.6) (Table 7).
We compared the clinical and laboratory parameters from a group of patients with AOSD with a large group of patients with infectious, malignant, and inflammatory diseases, all of them differential diagnoses in the setting of FUO. Even though we did not identify any single variable as pathognomonic of AOSD, we found that the presence of a cluster of clinical and laboratory parameters can reliably identify the majority of patients with AOSD.
The mean age of patients in the AOSD group (28 yr) was lower than the mean ages of patients in the control groups. In general, the duration of their disease was long (mean, 16 mo) and, as expected, they required lengthy diagnostic workups before the diagnosis was established. None of the previously published sets of criteria was used to diagnose AOSD. All the patients were extensively studied (17% underwent a diagnostic laparotomy), and AOSD was diagnosed in a conventional manner, after the exclusion of infectious, malignant, and inflammatory conditions, along with the consensus of rheumatologists. The accuracy of the diagnosis was high: all the patients remained as AOSD after a long follow-up (mean, 39.6 mo). The frequency of clinical and laboratory parameters did not differ from series reported elsewhere3,4,10,13,17.
The multivariate analysis identified 4 variables strongly associated with the presence of AOSD: arthritis, pharyngitis, splenomegaly, and neutrophilia. We incorporated them, along with Still rash, into a clinical scale meant to identify the patients with AOSD. Each parameter was assigned a different weight, according to the magnitude of its association with the presence of AOSD. We assessed the scale in our cases and controls, where it demonstrated a good accuracy, with a very high specificity and positive and negative predictive values higher than 90%. Accordingly, likelihood ratios, especially in case of a negative result, were very good. The latter is because the cutoff point of the scale was set high, in order to gain specificity (and a low negative likelihood ratio). We believe that by doing so the scale becomes very useful. Although it does not detect all the patients with AOSD, it pinpoints those who can be diagnosed easily, without further study. Those patients with AOSD who are not identified by the scale (roughly 20%) are those whose disease is not typical, cases in which other conditions must be sought and excluded before the AOSD diagnosis can be established. In light of these findings, we believe that the issue is not which diseases must be ruled out before establishing the diagnosis of AOSD, but rather, in which patients with FUO can AOSD be diagnosed before ruling out differential alternatives.
Seven studies have published sets of criteria for the classification of AOSD2-6,14,17. However, they have noteworthy limitations (Table 8). They have retrospective designs, and none has been prospectively validated. Only 2 studies, by Yamaguchi et al17 and Fautrel et al4, included patients with different diseases as controls. The Yamaguchi criteria require the exclusion of the differential diagnosis before the diagnosis is established. Thus, per se they are not able to distinguish between AOSD and similar conditions. The major limitation of the study by Fautrel and coworkers is that it was created from a database of patients for whom ferritin and glycosylated ferritin had been ordered. Thus, a selection bias is clearly present, and the results are hard to extrapolate to the clinical scenario. Furthermore, Fautrel's criteria require the quantification of glycosylated ferritin, a test not available in most centers.
The present work was based on a large, diverse group of patients undergoing a diagnostic procedure for FUO. Therefore its setting resembles very closely the clinical scenario in which the diagnosis of AOSD is most common and problematic. Another advantage of the study and of the proposed clinical scale is that it is simple. It relies on clinical findings and a routine laboratory test. The score can be calculated for a patient as soon as he or she is admitted. The scale provides a continuous variable and not a categorical answer; it conceptually resembles the clinical judgment used when making clinical decisions.
A major limitation of our study is its retrospective design. Due to the methodologic constraint, we could not include ferritin and glycosylated ferritin in the analysis. The retrospective quality of the study is difficult to avoid when dealing with uncommon conditions such as AOSD. However, it creates the need for prospective validation in order to prove the predictive value of the clinical scale. Perhaps the most important limitation of the study is that the same database was used to generate and to test the clinical scale. Once more, such drawback will have to be dealt with in future work.
Our work provides noteworthy information regarding FUO. It shows that the magnitude of the fever is a nonspecific finding not associated with any particular condition. Lymphopenia, classically associated with systemic lupus erythematosus, was present in patients with infectious, malignant, and autoimmune diseases. Conversely, although LDH elevation was present in all the groups, the magnitude of the abnormality was larger in patients with AOSD and malignant diseases, particularly non-Hodgkin lymphoma.
In the present study, we included a group of patients with FUO in whom the diagnostic workup yielded no diagnosis (Group 4). Their mean age was 33.5 years, and the mean duration of their disease was less than a year. Most of their accompanying complaints were imprecise. Although a large proportion of them complained of arthralgia (70%), arthritis was detected in only 30%. This fact underscores the importance of the physical examination, and suggests that joint inflammation is a much more solid and objective finding than noninflammatory joint pain. It is feasible to think these patients had benign, self-limiting conditions such as viral infections.
In summary, in the current study we describe the clinical characteristics of a series of patients with AOSD and compare them with a large group of patients with infectious, malignant, and inflammatory conditions in the setting of FUO. We suggest a clinical scale designed to identify the patients with AOSD who can be safely diagnosed without a long and costly diagnostic workup.
The authors thank Dr. Hilda Fragoso for technical assistance and Dr. Sergio Ponce de Leon for technical assistance and help with the statistical analyses.