Patient and clinical characteristics
Fifty patients account for a total 185 times of clinical episodes were included in the study (Table 1). Among these 50 patients, 7 (14%) were male and 43 (86%) were female, and the mean age at enrollment was 13.9 ± 4.4 years old. These 185 episodes were divided into 4 groups: infected-active episodes as group A (n = 102; 55%), infected-inactive episodes as group B (n = 11; 6%), noninfected-active episodes as group C (n = 59; 32%), and noninfected-inactive episodes as group D (n = 13; 7%) (Table 2). In the similar studies of adult SLE patients comparing parameters from the four groups, the trend of our CRP results is the same as three reported literatures [18, 28, 32]. While the trend of ESR and PCT are also in consistent with those of Schäfer et al. and Wang et al., respectively (supplement Fig. 1). Like the result from Broca-Garcia et al., infected-active group (group A) had the highest values of PLR among all four groups [28]. Our data is in consistent with Aringer et al. stating that, without infection, CRP levels are higher in active SLE than in inactive SLE, the opposite is true in infections in SLE patients [18, 37].
Table 2
Comparison 185 clinical episodes related to activity flare and/or acute infection.
Group (n = 185) | infected-active (n = 102) | infected-inactive (n = 11) | noninfected-active (n = 59) | noninfected-inactive (n = 13) |
LABORATORY PARAMETERS (mean ± SD) |
WBC (/uL) | 8430 ± 5807 | 6702 ± 3717 | 8441 ± 5793 | 10715 ± 5313 |
Hb (g/dL) | 9.2 ± 1.9 | 11.4 ± 2.3 | 10.7 ± 2.0 | 11.9 ± 1.9 |
PLT (x10^3/uL) | 176.8 ± 114.6 | 224.5 ± 86.0 | 229.5 ± 135.8 | 310.6 ± 100.6 |
Segment (%) | 77.7 ± 17.8 | 77.6 ± 9.5 | 72.8 ± 14.3 | 81.2 ± 7.5 |
Lymphocyte (%) | 12.9 ± 11.6 | 16.3 ± 8.4 | 19.9 ± 12.7 | 12.3 ± 5.2 |
CRP (mg/dL) | 2.25 ± 4.08 | 2.32 ± 2.27 | 1.02 ± 2.33 | 0.31 ± 0.42 |
ESR (mm/hr) | 41.0 ± 38.4 | 38.8 ± 18.6 | 33.5 ± 32.3 | 29.7 ± 19.5 |
ESR/CRP | 168.7 ± 433.6 | 94.2 ± 178.5 | 254.5 ± 387.0 | 271.5 ± 366.0 |
Procalcitonin (ng/mL) | 1.998 ± 6.232 | 0.54 ± 1.18 | 0.27 ± 0.44 | 0.06 ± 0.04 |
ANA titer | 1,303 ± 940 | 352 ± 476 | 1,187 ± 966 | 467 ± 578 |
C3 (mg/dL) | 62.71 ± 31.45 | 78.18 ± 30.50 | 58.96 ± 37.94 | 103.41 ± 24.75 |
C4 (mg/dL) | 13.67 ± 10.09 | 18.90 ± 10.85 | 11.23 ± 9.37 | 13.18 ± 6.17 |
anti-dsDNA (IU/mL) | 545.18 ± 492.99 | 601.38 ± 330.58 | 770.89 ± 418.78 | 291.57 ± 310.66 |
NLR | 22.19 ± 37.09 | 8.07 ± 7.66 | 9.47 ± 19.40 | 8.73 ± 5.40 |
PLR | 421.0 ± 396.5 | 302.0 ± 128.8 | 253.1 ± 177.2 | 305.0 ± 98.1 |
RPR | 0.15 ± 0.14 | 0.07 ± 0.02 | 0.11 ± 0.13 | 0.06 ± 0.03 |
CLINICAL SCORING INDICES |
SLEDAI-2K (mean ± SD) | 19.68 ± 8.30 | 9.64 ± 5.53 | 15.53 ± 8.29 | 10.15 ± 5.29 |
mild activity ( < = 6), (n, %) | 4 (7%) | 4 (36%) | 9 (15%) | 5 (38%) |
moderate activity (7–12), (n, %) | 17 (19%) | 3 (28%) | 17 (29%) | 4 (31%) |
severe activity (> 12), (n, %) | 76 (74%) | 4 (36%) | 33 (56%) | 4 (31%) |
Renal column (mean) | 9.81 | 3.64 | 5.69 | 7.69 |
CNS column (mean) | 2.78 | 1.09 | 0.68 | 0 |
Vasculitis column (mean) | 0.86 | 0 | 1.49 | 0.31 |
Arthritis column (mean) | 0.71 | 0.36 | 1.42 | 0.31 |
Myositis column (mean) | 0.24 | 0 | 0 | 0 |
Cutaneous column (mean) | 0.85 | 1.27 | 1.83 | 0.62 |
Serositis column (mean) | 1.02 | 0 | 0.51 | 0 |
Complement column (mean) | 1.62 | 0.73 | 1.66 | 0.77 |
SLICC/ACR Damage Index (SDI) (mean, range) | 3.01 ( 0–7) | 0.64 ( 0–2) | 0.69 ( 0–5) | 0.92 ( 0–3) |
Parameters predictive of activity flare
Among all the parameters analyzed, we found SDI score, SLEDAI 2K score, NLR, RDW-to-platelet ratio (RPR), ANA levels, anti-dsDNA levels, antiphospholipid Ab levels, and urine cast were positive predictors of activity flare. But that Hb levels, platelet levels, C3 levels, C4 levels, and urine nitrate were negatively associated with the occurrence of disease activity (supplementary Table 1a). Multivariate GEE analysis showed that SDI score, SLEDAI 2K score, NLR, Hb levels, platelet levels, RPR, and C3 levels were independent parameters for predicting SLE activity flare (Table 3a). Here we confirmed that NLR, PLR, and RPR is a useful marker for the assessment of disease activity in pediatric SLE patients [23, 29]. Combination of these seven parameters resulted in a model with calculated AUC of 0.8964 and with sensitivity of 82.2% and specificity of 90.9% (Fig. 1a).
Table 3
a: Multivariate GEE for outcome of activity flare. Seven significant effectors are shown.
Parameter | Estimate | Standard Error | 95% Confidence Limits | P-value |
SDI | -0.0148 | 0.0088 | -0.0320 | 0.0025 | 0.0936 |
SLEDAI 2K | 0.0108 | 0.0029 | 0.0050 | 0.0165 | 0.0002 |
NLR | 0.0013 | 0.0004 | 0.0005 | 0.0021 | 0.0010 |
Hb | -0.0305 | 0.0126 | -0.0552 | -0.0057 | 0.0159 |
PLT* | -0.0147 | 0.0265 | -0.0667 | 0.0373 | 0.5797 |
RPR | 0.1614 | 0.2026 | -0.2357 | 0.5585 | 0.4257 |
C3 | -0.0025 | 0.0013 | -0.0050 | -0.0001 | 0.0449 |
*: original values divided by 100. |
NLR: neutrophil-to-lymphocyte ratio; RPR: RDW-to-platelet ratio. |
Table 3
b: Multivariate GEE for outcome of acute infection. Eight significant effectors are shown.
Parameter | Estimate | Standard Error | 95% Confidence Limits | P-value |
SDI | 0.0782 | 0.0169 | 0.0451 | 0.1114 | < .0001 |
Fever temperature | 0.0997 | 0.0947 | -0.0859 | 0.2853 | 0.2926 |
CRP | 0.0341 | 0.0075 | 0.0195 | 0.0487 | < .0001 |
PCT | 0.0005 | 0.002 | -0.0035 | 0.0045 | 0.8069 |
Lymphocyte percentage | -0.0006 | 0.0034 | -0.0073 | 0.0061 | 0.8520 |
NLR | 0.0005 | 0.0007 | -0.0009 | 0.0019 | 0.4840 |
Hb | -0.0125 | 0.0243 | -0.0601 | 0.0352 | 0.6086 |
SLEDAI 2K renal score | 0.0086 | 0.0079 | -0.0069 | 0.0242 | 0.2769 |
NLR: neutrophil-to-lymphocyte ratio; PCT: procalcitonin. |
We then propose an Activity Predict Score formula:
Activity Predict Score = 1.1707–0.0146⋅SDI score + 0.0108⋅SLEDAI 2K score + 0.0013⋅NLR − 0.0305⋅Hb − 0.0147⋅PLT (original value divided by 100) + 0.1614⋅RPR − 0.0025⋅C3.
We obtained the largest Youden Index when the cut-off point of Activity Predict Score is cut at 0.76652; that is, when Activity Predict Score is greater than 0.76652, it will be classified to be an activity flare; if it is less than 0.76652, it will be classified as without activity flare.
Parameters predictive of acute infection
Using GEE, we found acute infection were associated with SDI score, fever temperature, CRP levels, PCT levels, NLR, PLR, and renal score of SLEDAI 2K while lymphocyte percentage, Hb levels, and urine nitrate were negative predictors of infectious events (supplementary Table 1b). Multivariate GEE analysis showed that SDI score, fever temperature, CRP level, PCT levels, lymphocyte percentage, NLR, Hb levels, and SLEDAI 2K renal score are independent parameters for predicting acute infection in SLE patients (Table 3b). Renal disease, despite being associated with infections in the univariate analysis, has not retained the statistical significance in the multivariate analysis in some series [38, 39]. However, our result confirmed that renal involvement is significantly associated with active infection in the multivariate GEE analysis, in consistent to previous study [40]. Of note, our data is in consistent with Pimentel et al. that any increase in the SDI was associated with the occurrence of serious infections [39]. We also showed that compared to PCT, CRP is a more sensitive and specific marker for diagnosing bacterial infections in SLE [41]. However, there had also been some reports that PCT is more specific and has better diagnostic accuracy than CRP for infection in SLE [15, 33, 42]. Combination of these eight parameters resulted in a model with calculated AUC of 0.7886 and with sensitivity of 63.5% and specificity of 89.2% (Fig. 1b).
Predicted by multiple GEE result, we also obtained Infection Predict Score:
Infection Predict Score = 0.4193 + 0.0782⋅SDI + 0.0997⋅fever temperature + 0.0341⋅CRP + 0.0005⋅PCT − 0.0006⋅Lymphocyte percentage + 0.0005⋅NLR − 0.0125⋅Hb + 0.0086⋅ SLEDAI 2K renal score.
When cut at a value of 0.58286, there will be the largest Youden Index; that is, when the Infection Predict Score is greater than 0.58286, it will be classified as acute infection; if it is less than 0.58286, it will be classified as without acute infection.
Development of a calculator model to simultaneously differentiate flares from infections
Multinomial logistic regression which described the probability of being in a specific group was used to analyze the individual effects of covariates (independent variables) on the discrete nominal outcomes. We select a total of 10 variables (SDI, SLEDAI 2K, fever temperature, PCT, lymphocyte percentage, NLR, Hb, PLT, RPR, C3) to establish multinomial logistic regression. The regression formula obtained by multinomial logistic regression is as follows:
By inputing the value of the selected parameters into these three equations, we can get the ratios value of πA, πB, πC and πD, respectively. If the value obtained is larger, the probability of being classified into that nominal group is greater (the group divided into D is the reference group), then we will classify particular episode into that group (group A, B, C). That is, if calculated ln(πA/πD) was greater than 1 and was the biggest number compared with others (ln(πB/πD) and ln(πC/πD)), the patient was categorized as group A. If all three calculated numbers [ln(πA/πD), ln(πB/πD), and In(πC/πD)] were below 1, the patient was categorized as group D. With combination of these ten parameters, we can simultaneously predict four groups with accuracy of 70.13% for infected-active group, 10% for infected-inactive group, 59.57% for noninfected-active group, and 84.62% for noninfected-inactive group, respectively. From our multinomial logistic regression analysis, we identified SDI, SLEDAI 2K, fever temparature, PCT, lymphocyte percentage, NLR, Hb, PLT, RPR, and C3 as influencing factors for simultaneously differentiating activity flares from acute infections.
Evaluation of possible associated interaction between acute infection and activity flare
In order to observe whether there is an associated interaction between acute infection and activity flare, we intend to compare parameters from combined groups with or without infection (Fig. 2). We found that the levels of CRP, PCT, lymphocyte percentage, NLR, PLR, SLEDAI 2K and SDI from combined group with infection were significantly higher than those of combined groups without infection. But the levels of ESR, C3 and C4 are not significantly different between the two combined groups. Our results (elevated levels of SLEDAI 2K, SDI, NLR, and PLR under noninfected condition) indicate that acute infections might play the trigger role of activity flare. On the other hand, for those protein participate in both SLE disease inflammation and acute phase inflammation, there is no significant difference in their levels of ESR, C3 and C4 with or without regarding infections.
Trend analysis of parameters changes with time through hospitalization
The results of their mean baseline level and changes per time interval in different groups are shown in supplement Table 2. There significant differences for ESR, NLR, lymphocyte percentage, C3, C4, as shown in Fig. 3. From Fig. 3 (a) we found that ESR decreased with time, but the trend of decreasing was more prominent in group A than in group C. ESR appears to be a useful biomarker for SLE activity assessment. An elevated ESR is included in three out of five validated SLE activity scores [37]. The reported trend of ESR from Dima et al. are similar with our trend analysis indicating that the higher initial ESR level reflect the effect of both activity and infections in group A. From the trend difference between group A and C, we could differentiate noninfected-active SLE episodes (group C) from infected-active SLE episodes (group A) by the change patterns of ESR, NLR, lymphocyte percentage, C3, and C4 over time.