Demographic and clinical characteristics
2869 laboratory confirmed COVID-19 inpatients were screened, and 162 patients tested with autoimmunological detections were enrolled. Of the enrolled patients, 120 patients were detected with ANA+ENA (36% positive rate), 75 patients were detected with ACA (32% positive rate), 114 and 110 patients were detected with RF and ASO with 8% positive rate for each, respectively (Figure 1). All positive results of autoimmunological detections were newly emerging after SARS-CoV-2 infection compared with patient’s most recent result if any. There were 105 patients in the control group “Non-severe”, and 57 patients in the case group “Severe” (Supplementary Table 1). Elderliness (older than 65) do exhibit significant risk for severity (p< 0.001), both univariable and multivariable logistic regression analyses indicate patients aged over 65 are more vulnerable to severe progression than those aged less than 45 (p= 0.002 for multivariable analysis, hereinafter the same), however, those aged between 45 and 65 suffer the most (p< 0.001). Sex (p= 0.393) and the onset of symptoms to treatment time (OTT) (p=0.106) seem to contribute no significant risk. Comorbidity do serve as risk factor (p= 0.004), hypertension (p= 0.003), diabetes (p= 0.003) or coronary heart disease (p= 0.001) contributes significant risk. As a treatment option, severe patients are more often administered corticosteroid or immunoglobulin (p< 0.001). Before the stratification analysis with antinuclear autoimmunity, the overall profiles of the enrolled patients are in accord with those of other large-scale study3, indicating the representativeness of the study objects.
Antinuclear autoimmunity and severity of COVID-19
Of the detected patients, ANA+ENA is shown significant correlation with severity overall (p< 0.001). For specific single item, ANA IgG and anti-SSA account for 22% and 13% positive rate, respectively, and contribute significantly (p< 0.001 for each, respectively), ACA detection is shown significant but inverse correlation with severity for only IgG antibody (p= 0.038). RF and ASO are demonstrated no correlations with severity (p= 0.752 and p= 0.999, respectively). (Supplementary Table 2)
Univariable and multivariable logistic regression analyses were performed to reveal the correlations between severity and multiple potential risk factors including ANA+ENA detection, age, sex, OTT and comorbidity (only risky chronic diseases included as hypertension or diabetes or CHD or COPD) (Table 1). Compared with model without ANA+ENA detection shown in Supplementary Table 1, ANA+ENA detection reflecting antinuclear autoimmunity was taken into account here and demonstrated significant correlation with severity in both univariable and multivariable analyses, and even higher risk in multivariable analysis (OR= 10.514, 95%CI 3.633- 30.424, p< 0.001). With the multivariable analysis involving antinuclear autoimmunity compared with univariable analysis, age over 65 loses the correlation with severity (p= 0.064), however, the middle age (45≤age≤65) (p= 0.011) and comorbidity (p= 0.002) are still strongly correlated, and interestingly, OTT becomes correlated instead, indicating antinuclear autoimmunity involvement attenuates the risk of elderliness and manifests the urgency of timely treatment. Sex is not correlated for either univariable or multivariable analysis.
Table 1
Potential risk factors including antinuclear autoimmunity for severe progression of COVID-19
|
Total
(n=120)
|
Non-severe
(n=74)
|
Severe
(n=46)
|
Univariable OR
(95% CI)
|
p value
|
Multivariable OR
(95% CI)
|
p value
|
ANA+ENA
|
|
|
|
|
|
|
|
Positive
|
43 (36%)
|
15 (20%)
|
28 (61%)
|
6.119
(2.696- 13.887)
|
<0.001*
|
10.514
(3.633- 30.424)
|
<0.001*
|
Negative
|
77 (64%)
|
59 (80%)
|
18 (39%)
|
Age
|
|
|
|
|
|
|
|
>65
|
36 (30%)
|
13 (18%)
|
23 (50%)
|
3.073
(1.270- 7.437)
|
0.013*
|
..
|
0.064
|
45≤Age≤65
|
52 (43%)
|
33 (44%)
|
19 (41%)
|
12.385
(3.552- 43.185)
|
<0.001*
|
7.235
(1.579- 33.140)
|
0.011*
|
<45
|
32 (27%)
|
28 (38%)
|
4 (9%)
|
1.000
|
<0.001*
|
1.000
|
0.030*
|
Sex
|
|
|
|
|
|
|
|
Male
|
49 (41%)
|
27 (36%)
|
22 (48%)
|
..
|
0.219
|
..
|
0.051
|
Female
|
71 (59%)
|
47 (64%)
|
24 (52%)
|
OTT, days
|
|
|
|
|
|
|
|
>7
|
67 (56%)
|
37 (50%)
|
30 (65%)
|
..
|
0.103
|
2.924
(1.041- 8.212)
|
0.042*
|
≤7
|
53 (44%)
|
37 (50%)
|
16 (35%)
|
Comorbidity (Hypertension or Diabetes or CHD or COPD)
|
|
|
|
With
|
38 (32%)
|
12 (16%)
|
26 (57%)
|
6.717
(2.872- 15.709)
|
<0.001*
|
5.256
(1.805- 15.306)
|
0.002*
|
Without
|
82 (68%)
|
62 (84%)
|
20 (43%)
|
Data are n (%), or n/N (%), unless otherwise specified; OTT= onset of symptoms to treatment time (days); CHD= coronary heart disease; COPD= chronic obstructive pulmonary disease. Univariable or multivariable logistics regression analysis was used to compare differences between control group “Non-severe” and case group “Severe”. p< 0.05 was taken as statistically significant. |
Since strongly correlated with severity, ANA+ENA detection was further subjected to hierarchical cluster analysis, and ANA+ENA is overall shown to possess 60.9% sensitivity and 79.7% specificity with moderate correlation when indicating severity (Pearson R= 0.412, OR= 6.119, 95% CI 2.696- 13.887, p< 0.001) (Table 2). It becomes further correlated for cluster middle-aged patients (45≤age≤65), the sensitivity and specificity are elevated to 78.9% and 84.8%, respectively, with strong correlation (Pearson R= 0.631, OR= 21.000, 95% CI 4.893- 90.126, p< 0.001), however not significantly correlated for patients aged over 65 (p= 0.452) and less than 45 (p= 0.057) (Table 2). For cluster female patients, stronger correlation with 79.2% sensitivity and 80.9% specificity between ANA+ENA and disease severity is observed (Pearson R= 0.581, OR= 16.044, 95% CI 4.717- 54.568, p< 0.001), however not for the males (p= 0.158). When cluster narrowed to middle-aged female, strongest correlation with 100.0% sensitivity and 81.0% specificity is observed (Pearson R= 0.770, p< 0.001). The time from onset of symptoms to treatment (OTT, days) is thought to reflect illness urgency, and underlying chronic comorbidities make patients vulnerable to severe progression, in this study, no matter OTT longer or shorter than 7 days, with or without pre-existing comorbidity, the ANA+ENA detection can both indicate severity significantly.
Table 2
Hierarchical cluster analyses of correlations between ANA+ENA detection and severity of COVID-19
Variables
|
ANA+ENA
|
Severe
(n)
|
Non-severe
(n)
|
Sensitivity/ Specificity (%/ %)
|
Pearson
Ra
|
p value
|
OR
(95% CI)
|
Mantel-Haenszel
Common ORb (95% CI), p value
|
Overall
|
+
|
28
|
15
|
60.9%/ 79.7%
|
0.412
|
<0.001*
|
6.119
(2.696- 13.887)
|
..
|
-
|
18
|
59
|
Age, years
|
|
|
|
|
|
|
|
|
|
>65
|
+
|
10
|
4
|
43.5%/ 69.2%
|
..
|
0.452
|
..
|
..
|
|
-
|
13
|
9
|
|
45≤Age≤65
|
+
|
15
|
5
|
78.9%/ 84.8%
|
0.631
|
<0.001*
|
21.000
(4.893- 90.126)
|
..
|
|
-
|
4
|
28
|
|
< 45
|
+
|
3
|
6
|
75.0%/ 78.6%
|
..
|
0.057
|
..
|
..
|
|
-
|
1
|
22
|
Sex
|
|
|
|
|
|
|
|
|
|
Male
|
+
|
9
|
6
|
40.9%/ 77.8%
|
..
|
0.158
|
..
|
..
|
|
-
|
13
|
21
|
|
Female
|
+
|
19
|
9
|
79.2%/ 80.9%
|
0.581
|
<0.001*
|
16.044
(4.717- 54.568)
|
|
-
|
5
|
38
|
Female and 45≤Age≤65
|
|
|
|
|
|
|
|
Yes
|
+
|
11
|
4
|
100.0%/ 81.0%
|
0.770
|
<0.001*
|
..
|
..
|
-
|
0
|
17
|
No
|
+
|
17
|
11
|
48.6%/ 79.2%
|
0.292
|
0.006*
|
3.606
(1.411- 9.214)
|
-
|
18
|
42
|
OTT, days
|
|
|
|
|
|
|
|
|
|
>7
|
+
|
16
|
6
|
53.3%/ 83.8%
|
0.393
|
0.001*
|
5.905
(1.906- 18.293)
|
7.110
(2.986- 16.926), p< 0.001*
|
|
-
|
14
|
31
|
|
≤7
|
+
|
12
|
9
|
75.0%/ 75.7%
|
0.476
|
0.001*
|
9.333
(2.400- 36.296)
|
|
-
|
4
|
28
|
Comorbidity (Hypertension or Diabetes or CHD or COPD)
|
|
|
|
|
|
|
With
|
+
|
15
|
1
|
57.7%/ 91.7%
|
0.465
|
0.004*
|
15.000
(1.679- 134.025)
|
8.051
(3.054-21.222),
p< 0.001*
|
|
-
|
11
|
11
|
|
Without
|
+
|
13
|
14
|
65.0%/ 77.4%
|
0.388
|
<0.001*
|
6.367
(2.130-19.030)
|
ANA= antinuclear antibody, ENA= extractable nuclear antigen, OTT= onset of symptoms to treatment time (days), OR= odds ratio. *Cochran-Mantel-Haenszel test (CMH test), X2
test, or Fisher’s exact test was used to compare differences between control group “Non-severe” and case group “Severe” where appropriate. p<0.05 was taken as statistically significant.
aPearson correlation coefficient R: “very weak” 0.00- 0.19, “weak” 0.20- 0.39, “moderate” 0.40-0.59, “strong” 0.60-0.79, “very strong” 0.80- 1.00.
bIf Homogeneity of OR was tested p>0.05, Mantel-Haenszel Common OR estimate was applied.
|
For specific single item in the detection list of ANA+ENA, ANA IgG or anti-SSA is significantly correlated with severity of COVID-19 separately (p< 0.001 for each, respectively) (Supplementary Table 2) with high specificity (93.2% and 95.9%, respectively), but low sensitivity (45.7% and 28.3%, respectively) (Supplementary Table 3 and supplementary Table 4). Among anti-SSA antibody positive cases, 13 present with anti-Ro-52 (81.3%), and 7 present with anti-Ro-60 (43.8%) (data not shown).
Demographic and clinical characteristics between cluster ANA+ENA positive and negative are compared and shown in Figure 2A. Compared with cluster negative, the complication myocardial injury (28%, p= 0.007) and severe cases (65%, p< 0.001) present significantly more frequently in the cluster ANA+ENA positive. Sex, age and some clinical characteristics such as OTT, comorbidity, and hepatic or renal injury complication, all exhibit no difference between the two clusters. However, when severe cases are picked out for hierarchical cluster analyses, female (68%, p= 0.008) and middle-aged patients (45≤age≤65) (54%, p= 0.035) present more frequently in the cluster ANA+ENA positive, and so does the severe middle-aged female patients (39%, p= 0.041), none of this kind of patient is found in the cluster negative in this study. It seems more patients of positive ANA+ENA spent shorter OTT but statistically insignificant (p= 0.152). For these two types of severe clusters, ANA+ENA positive and negative, the incidence of hepatic, renal or myocardial injury seems to be of no difference (Figure 2B). Incidence rates of symptoms between cluster ANA+ENA positive and negative are shown in Figure 2C. Incidence of fever and chest distress are statistically more frequent in cluster positive (84%, p= 0.007, and 70%, p= 0.015, respectively) compared with cluster negative (60% and 47%, respectively), and there is no statistical difference for other symptoms. However, when severe cases picked out, the incidences of all symptoms become statistically indistinguishable between the two clusters (Figure 2D).
In order to investigate if there is difference of the time-dependent progression of disease for severe cases between cluster ANA+ENA positive and negative, the timelines of different events are compared and demonstrated in Figure 3A and Figure 3B. The onset of symptoms to severity time (OTS, days) of the cluster positive, of which 89% is shorter than 12 days (p< 0.001) (Figure 3E), appears to be significantly shorter than that of cluster negative (p= 0.021) (Figure 3C). However, the overall disease courses, namely, the onset of symptoms to outcome time (OTO, days) are statistically the same for these two clusters (p= 0.735) (Figure 3D). Interestingly, during treatment, 7 severe patients experienced post-remission aggravation, 6 of them (21%) present in the cluster positive, and only 1 (6%) presents in the cluster negative (Figure 3E), although the difference is not supported by statistics (p= 0.220). The incidence of hepatic, renal or myocardial injury appears no statistical difference between the two clusters (Figure 2B, Figure 3A and Figure 3B). 3 patients died, all are senile (all aged 75), 2 in cluster negative and 1 in cluster positive who advanced quickly to death with only 12 days overall disease course (Figure 3A).
Disease dynamics picturing
Disease dynamics picturing of antinuclear autoimmunity involved COVID-19 was profiled by a representative case study. A 49-year-old woman was admitted on Feb 6, with 2-day history of fever, fatigue and intermittent dry cough, SARS-CoV-2 infection was confirmed by RT-PCR test of pharyngeal swab specimen. She stated 10-year hypertension history but denied any autoimmune diseases or other comorbidities. The thoracic CT imaging exhibited no sign of COVID-19 on Feb 5, but started to show multiple ground-glass opacities on Feb 9, especially in subpleural area. She progressed rapidly to acute respiratory distress syndrome (ARDS) and was transferred to ICU immediately on Feb 12. The ARDS progressed further and reached peak on Feb 14, ventilator was applied promptly. With intensive care and massive use of corticosteroid (total 2500 mg), immunoglobulin, antibiotics and other essential treatments, she gradually recovered with RNA turning negative since Feb 21 with a solo positive result in the middle on Feb 27, and eventually was discharged on Mar 11 (Figure 4, schemes are to scale).
The patient was detected presence of multiple serous antinuclear antibodies with positive ANA IgG, anti-SSA/Ro-52 and anti-AMA-M2 on Feb 10, which was only 6 days after onset of symptoms, 2 days before the initiation and 4 days before the peak of ARDS. The timing was as early as other sensitive inflammation indicators such as Interlukin-6 (IL-6) and Serum Amyloid A (SAA) which reached peaks as early as Feb 10 with the amount of 12.37 folds and 12.63 folds of ULN (upper limit of normal), respectively. C-reactive protein (CRP) and D-dimer followed, both reaching peaks on Feb 15 with 17.03 folds and 9.63 folds of ULN, respectively. Lactic dehydrogenase (LDH) acted much slower, peaking on Feb 20 with 1.85 folds of ULN. As myocardial damage indicators, creatine kinase-MB (CK-MB) elevated preceding to CK, peaking on Feb 20 with 2.00 folds of ULN, 2 days before CK’s. The antinuclear antibodies still stayed positive with ANA IgG, anti-SSA/Ro-52, anti-SSA/Ro-60 and anti-PR3-ANCA on Feb 24 when a post-remission aggravation of disease occurred, and all turned negative on Mar 9 when disease recovered. Since the onset of symptoms, lymphocytes had kept dropping until Feb 15 reaching nadir (0.23 folds of LLN (lower limit of normal)), coincident with the ARDS peak. Eventually, most indicators went down to around baseline except SAA when disease recovered (Figure 4, Supplementary table 5).