Abstract
Objective. To determine the value of clinical measures in diagnosis of adult-onset Still’s disease (AOSD), and to identify the optimal set of proposed classification criteria, in a Chinese population.
Methods. A total of 70 patients with AOSD and 140 non-AOSD inpatients with fever were retrospectively identified at Zhongshan Hospital, Shanghai, from January 2003 to December 2009. Clinical measures and 4 sets of diagnostic criteria (Yamaguchi, Calabro, Cush, and Reginato) were evaluated by sensitivity, specificity, positive/negative predictive value (PPV, NPV), and positive/negative likelihood ratio (PLR, NLR) for diagnosis of AOSD.
Results. In our series, higher sensitivity included hyperpyrexia (temperature ≥ 39°C, 94.29%), arthralgia (80.0%), polymorphonuclear neutrophils (PMN) ≥ 75% (84.29%), serum ferritin ≥ 2-fold the upper normal value (90.0%), negative antinuclear antibodies (85.29%), and rheumatoid factor (84.38%); while higher specificity included transient erythema (98.57%), sore throat (85.0%), leukocytes ≥ 15,000/mm3 (87.86%), and PMN ≥ 85% (85.0%). Rash, arthralgia, and sore throat were found to have better sensitivity and specificity (PLR 3.29–4.86). Leukocytes ≥ 10,000/mm3, PMN ≥ 80%, and serum ferritin ≥ 5-fold the upper normal limit were set as critical points. The Reginato criteria set had the highest specificity, 99.29%. The Yamaguchi set had the highest sensitivity, 78.57%, with a better accuracy of 87.14%.
Conclusion. The Yamaguchi diagnostic criteria had better accuracy in Chinese patients. Indicators such as rash, arthralgia, sore throat, leukocytes ≥ 10,000/mm3, PMN ≥ 80%, and serum ferritin ≥ 5-fold the upper normal limit were helpful for diagnosis of AOSD. We recommend using these indicators in combination instead of alone.
Adult-onset Still’s disease (AOSD) is a rare systemic inflammatory disorder of unknown etiology. Clinical presentation is nonspecific and heterogeneous. Symptoms commonly include high-spiking fevers, evanescent rash, and arthritis/arthralgia. Multiple organs may be involved simultaneously. Since there is no specific feature, no consensus on diagnostic criteria has been reached. The most commonly used classification criteria1 were proposed by Yamaguchi, et al2,3,4,5,6,7,8,9,10,11,12,13. Other criteria include those of Cush, et al14, Calabro, et al15, and Reginato, et al16. All these criteria are based on elimination of other diseases, such as chronic infection, tumors, and other autoimmune disorders. Differential diagnosis is often based on treatment responses, and therefore may delay the establishment of the true causes. To date, few studies have examined the validity of the above-noted classification criteria17,18,19.
Our objective was to assess the influence of the various clinical and laboratory measures suggestive for AOSD, and determine a more accurate set of classification criteria for AOSD in a Chinese population.
MATERIALS AND METHODS
Patients
We retrospectively reviewed the records of 210 adult patients treated at the Department of Internal Medicine, Zhongshan Hospital (affiliated to Fudan University), from January 2003 to December 2009. Among them, 70 cases were labeled AOSD by the attending physicians; a diagnosis other than AOSD was established for the remaining 140 cases complaining of fever during the same period. Demographic and medical information were extracted from medical records and telephone followup, and entered into the database using a standardized questionnaire. Results of routine laboratory tests, such as hemograms, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), and liver enzymes, were available for all patients. The majority of cases also had data concerning antinuclear antibodies (ANA), rheumatoid factor (RF), and serum ferritin at the time of diagnosis.
Classification procedure
This was a retrospective analysis. The diagnosis of AOSD was based on the following steps: (1) The initial clinical assessment of patients presenting with clinical features (the typical symptoms and signs) of AOSD. (2) A thorough history, physical examination, and laboratory assessment for differential diagnosis; alternative diagnoses (e.g., infections, malignant tumor, and other autoimmune disorders) were fully considered by the attending physicians and eliminated based on an extensive panel of analysis such as bone marrow, imaging examinations, pathologic findings, and microbiological test results. ANA, anti-dsDNA antibody, and antineutrophil cytoplasmic antibody were routinely investigated to exclude other autoimmune diseases. (3) Assessment of prognosis: The “gold standard” diagnostic criteria were referred to for the effect of treatments, the disease courses, and complications. All patients were followed up at clinics or by telephone.
The mean followup period of patients with AOSD was 2.5 ± 1.6 years (range 1∼80.4 months). Sixty-eight patients (97.1%) received glucocorticoid therapy, except for 2 patients who had good response to nonsteroidal antiinflammatory drugs. Sixty patients (83.6%) exhibited a monocyclic disease pattern, 4 (6.6%) experienced a polycyclic pattern, 3 (4.9%) exhibited a chronic course and had persistent articular symptoms in the absence of systemic features, and 3 (4.9%) died.
The control group included 140 patients with diseases other than AOSD, including infectious, neoplastic, and autoimmune diseases, established on the basis of clinical manifestations, laboratory tests, imaging examinations, pathologic findings, and microbiological test results.
Statistical analysis
Chi-square testing was performed for comparison of proportions. When small numbers of patients were compared, Fisher’s exact test was used. Clinical features and laboratory tests were evaluated for sensitivity, specificity, positive/negative predictive value (PPV, NPV), and likelihood ratio of a positive/negative test result (PLR, NLR). Sensitivity and specificity were defined as the percentage of positive (sensitivity) or negative (specificity) findings of the variable from all AOSD or non-AOSD patients, according to the diagnostic gold standard. Positive and negative predictive values were defined as the percentage of the measure by correctly identified positives or negatives, respectively, from all positives or negatives. The PLR was calculated as (sensitivity/1 – specificity). The NLR was calculated as (1 – sensitivity/specificity). Receiver operating characteristic (ROC) curves of leukocytes, polymorphonuclear neutrophils (PMN), and serum ferritin were generated. A p value < 0.05 was considered statistically significant.
RESULTS
Demographics and control diagnosis
Seventy patients with AOSD (26 men, 44 women; age 36.1 ± 14.5 yrs, range 18–72 yrs, at time of diagnosis) were identified. Of the 140 control patients (75 men, 65 women; age 46.9 ± 18.5 yrs, range 16–89 yrs, at diagnosis), 69 had infectious diseases, 19 had neoplasms, and the remaining 52 had systemic autoimmune disorders other than AOSD (Table 1).
Clinical features
The most common symptoms and signs included fever (100%), hyperpyrexia (temperature ≥ 39°C, 94.29%), arthralgia (80.0%), sore throat (72.86%), lymphadenectasis (71.43%), rash (75.71%), and myalgia (41.43%). Compared with previous large series, fever (84.7%∼100%), arthralgia (64.1%∼100%), sore throat (38.7%∼91.9%), myalgia (56.2%∼83.9%), and rash (51.8%∼87.1%) were reported3,8,17,18,19.
In comparison to disease controls, the frequency of rash, arthralgia, arthroncus, sore throat, myalgia, lymphadenectasis, hepatomegaly, leukocytosis, neutrophilia, negative ANA and RF, and serum ferritin in AOSD patients was significantly higher (p < 0.05 for all). There were no significant differences in the frequency of cardiac effusion, pleural effusion, interstitial pneumonia, anemia, and thrombocytosis between the 2 groups.
Diagnostic efficacy of clinical and laboratory measures
Clinical features that were sensitive for establishing diagnosis of AOSD (≥ 80%) included hyperpyrexia (≥ 39°C, 94.29%) and arthralgia (80.0%) (Table 2). Clinical features with ≥ 80% specificity included transient erythema (98.57%) and sore throat (85.0%). The highest PPV was for transient erythema (92.59%). The top 4 highest NPV were hyperpyrexia (≥ 39°C, 92.86%), arthralgia (88.33%), rash, (86.72%) and sore throat (86.23%). Laboratory tests with ≥ 80% sensitivity included PMN ≥ 75% (84.29%), serum ferritin ≥ 2-fold the upper normal limit (90.0%), negative ANA (85.29%), and RF (84.38%). Laboratory tests with ≥ 80% specificity included leukocytes ≥ 15,000/mm3 (87.86%) and PMN ≥ 85% (85.0%). The measures with higher PPV were leukocytes ≥ 15,000/mm3 (67.92%) and serum ferritin ≥ 5-fold the upper normal limit (70.69%). The 3 laboratory tests with the highest NPV were leukocytes ≥ 10,000/mm3 (86.49%), PMN ≥ 75% (88.30%), and serum ferritin ≥ 4-fold the upper normal limit (86.96%). The PLR was > 2.5 for hyperpyrexia (≥ 40°C, 2.52), transient erythema (24.97), arthralgia (3.29), rash (3.65), sore throat (4.86), leukocytes ≥ 10,000/mm3 (2.50), PMN ≥ 80% (2.72), and serum ferritin ≥ 5-fold the upper normal limit (3.04). The highest PLR was for transient erythema, 24.97. The NLR was low for hyperpyrexia (≥ 39°C, 0.15), arthralgia (0.26), PMN ≥ 75% (0.26), and serum ferritin ≥ 2-fold the upper normal limit (0.20).
Serial tests revealed increased specificity and PLR when 1 out of the following 6 measures was combined with hyperpyrexia (≥ 39°C): rash, sore throat, arthralgia, leukocytes ≥ 10,000/mm3, and PMN ≥ 80% (76.43%∼95.71% vs 37.14%∼85.0% for single items; PLR 3.21∼12.99 vs 1.50∼4.86). The combination of rash with the following items had PLR > 6: leukocytes ≥ 10,000/mm3 (PLR 8.66), arthralgia (PLR 11.20), sore throat (PLR 7.80), and PMN ≥ 80% (PLR 6.17). A majority of the combination of 3 items yielded a PLR > 10: 28.97 for rash or sore throat or leukocytes ≥ 10,000/mm3; 24.03 for hyperpyrexia (≥ 39°C) or arthralgia or sore throat; and 20.03 for sore throat or arthralgia or leukocytes ≥ 10,000/mm3.
A comparison of our results with those from the Fautrel18 and Yamaguchi1 criteria is shown in Table 3.
Comparison of different classification criteria sets
Among the 4 classification criteria sets, the Reginato criteria had the highest specificity, 99.29%, with accuracy of 83.33%. The Yamaguchi criteria set had the highest sensitivity, 78.57%, with better accuracy: 87.14%. A comparison of our criteria series with those of Fautrel18 and Yamaguchi1 is shown in Table 4.
DISCUSSION
Fever is a prominent feature of AOSD (incidence 95.7%–100%) and frequently presents as hyperpyrexia (≥ 39°C)2,3,4,5,6,7,8. In our study, hyperpyrexia (≥ 39°C) was noted in 66 patients with AOSD (94.29%) and displayed a NPV of 92.86% with a NLR of 0.15, indicating that a diagnosis of AOSD is unlikely in patients without hyperpyrexia. However, fever is also the common symptom for tumors and infections. A previous study20 reported only slight fever in elderly patients with AOSD. Caution must be exercised to exclude AOSD if the patient has unexplained fever. Therefore patients with fever were selected as the control group in our series.
In our series, rash, arthralgia, and sore throat1,16,18,21 were more sensitive (> 70%) and specific (> 70%), with a PLR of 3.29–4.86, suggesting that these 3 features are helpful for establishing a diagnosis of AOSD. The NLR of the 3 features was lower (0.26–0.32), indicating their value in ruling out AOSD based on the absence of these features. Actually, articular involvement is included in all major AOSD criteria sets1,14,15,16. Rash is listed in the major criteria in most criteria sets1,15,16. Transient erythema had a specificity of 98.57% and a PLR of 24.97 in our series. Its sensitivity, however, was lower (35.71%), and therefore it cannot be used alone in a criteria set. Sore throat has been adopted only by the Yamaguchi and Reginato criteria as a minor item. We noted a higher PLR for sore throat than for any other clinical or laboratory measure except for transient erythema in our series. As a prominent symptom of AOSD, sore throat is also the common symptom in thyroid and upper respiratory tract disorders. Careful inquiry and physical examination are essential in order to establish a diagnosis of AOSD.
Leukocytosis and/or neutrophilia were added to all criteria sets, but the cutoff value was inconsistent. Better sensitivity (> 70%) and specificity (> 70%) with a PLR of 2.7 and NLR of 0.40 were observed when the cutoff value of leukocytes (PLR 2.5; NLR 0.31) was set as not less than 10,000/mm3 and PMN was 80% in the study (ROC of leukocytes and PMN are shown in Figure 1A and Figure 1B, respectively)1,16. Serum ferritin is an acute-phase reactant. A significant increase in serum ferritin has been proposed as an item of diagnosis or activity marker4,5,6,7,8,10,11,12,13. In our series, multiple cutoff values were used for serum ferritin, including > 2, > 4, and > 5-fold the upper normal limit. Among these cutoff values, an increase > 5-fold had the highest specificity at 73.02% (compared to 50.79% for > 2-fold) and a PLR of 3.04 (compared to 1.83 for > 2-fold); the sensitivity did not change significantly (the ROC for serum ferritin is shown in Figure 1C). We therefore recommend using serum ferritin ≥ 5-fold the upper normal limit as the critical point for diagnosis of AOSD12. Negative ANA and RF have high sensitivity (84.38%, 94.29%, respectively) but low specificity (30.0%, 37.14%, respectively), indicating these 2 measures are prone to misdiagnosis if used singly. Thus these 2 measures became one of the items in various criteria sets for differentiation of other autoimmune diseases1,14,15,16.
Our study also validated that the diagnostic possibility of AOSD could increase greatly when findings of hyperpyrexia were combined with 2 or 3 of sore throat, arthralgia, rash, PMN ≥ 80%, and leukocytes ≥ 10,000/mm3.
Among the 4 criteria sets included in our study, the Reginato criteria displayed the highest specificity, 99.29%; whereas the Yamaguchi criteria had the highest sensitivity, 78.57%, and better accuracy: 87.14%. The results from the Yamaguchi set are similar to the conclusions of Fautrel, et al18, and the results from the Reginato or Calabro sets are similar to Yamaguchi1. The sensitivities of various sets of AOSD criteria are lower than that of the study of Masson, et al17 (Reginato: 55.25%; Yamaguchi: 93.5%; Cush and Calabro: 80.6%).
The sensitivity and specificity of many AOSD disease features, including hyperpyrexia, arthralgia, and negative ANA and RF, in our series differed significantly in comparison with previous reports. For example, the specificity of fever in our study was lower than previously reported. Different control groups may have contributed to the observed differences. Our controls complaining of fever were different from those of the Fautrel study completing ferritin and glycosylated ferritin testing18, and were also different from the Yamaguchi study including rheumatic disease, infectious disease, and fever of unknown origin1. Moreover, the control group in our study included more disorders that are difficult to differentiate from AOSD (Table 1), and this may have contributed to different sensitivity and specificity.
It is noteworthy that the frequency of myalgia (41.43%) was lower in our series compared with that in other reports (56.2%–83.9%). Consistent with our findings, one study conducted in Chinese people reported a low frequency of myalgia (27.9%)10, thus indicating an ethnic difference.
In summary, the Yamaguchi diagnostic criteria1 were found to have better accuracy in our series. Our results also indicate that rash, arthralgia, sore throat, leukocytes ≥ 10,000/mm3, PMN ≥ 80%, and serum ferritin ≥ 5-fold the upper normal limit are helpful for diagnosis of AOSD. We recommend using these features in combination instead of alone.
- Accepted for publication November 30, 2010.