Demographic characteristics of cohorts
A total of 402 patients were included in this study, including 255 with AIS, 147 with other mild neurological diseases and without any vascular diseases as controls. Baseline demographic characteristics were presented in Table 1. Patients in AIS group were predominantly males (173, 63.8%), with an older mean age (70.2±13.3) and higher systolic (152.4±20.7) and diastolic (81.7±13.2) blood pressures than controls. In addition, the proportion of alcohol consumption (244,96.8%), use of antihypertensive drugs (157,94.0%), and lipid-lowering drugs (156,94.0%) were higher in the AIS group compared with the control group. The differences between the above indicator groups were statistically significant (p<0.05). BMI and smoking history were balanced and not statistically significant.
Table 1. Characteristics of study subjects
Demographics
|
AIS group (n = 255)
|
Control group (n = 147)
|
p-value
|
Sex male, n (%)
|
173 (67.8)
|
69 (46.9)
|
<0.001
|
Age, years, (IQR)
|
71 (62,81)
|
58 (41.25,67)
|
<0.001
|
Body mass index, (IQR)
|
23.49 (21.21,25.78)
|
25.95
(22.25,52.50)
|
0.546
|
Systolic BP (IQR), mmHg
|
152 (143.25,166)
|
118.50 (73,138)
|
<0.001
|
Diastolic BP (IQR), mmHg
|
80 (73,90)
|
82 (71,119.25)
|
<0.001
|
Having smoking history, n (%)
|
80 (31.7)
|
20 (21.3)
|
0.056
|
Having alcohol history, n (%)
|
244 (96.8)
|
11 (11.7)
|
<0.001
|
Using antihypertension drugs, n (%)
|
157 (94.0)
|
67 (84.8)
|
0.018
|
Using lipid-lowering drugs, n (%)
|
156 (94.0)
|
67 (84.8)
|
0.019
|
|
|
|
|
|
AIS = acute ischemic stroke; Systolic BP = systolic blood pressure; Diastolic BP = diastolic blood pressure
Survival analyses
To reflect the survival difference, the Kaplan-Meier curve was used to compare the survival probability of the AIS group with that of the control group, as shown in Figure 2A. It could be found that within 10 months, the survival probabilities of the two groups tended to be close to each other and remained around 1; then after 10 months, the survival probabilities of patients in the AIS group decreased significantly compared with the control group, and the difference was statistically significant (log-rank: p < 0.001); while the survival probabilities of patients in two groups tended to be equal in the next 30 months. To avoid gender bias as indicated by the differences between the two groups, a subgroup analysis was done to explore the survival differences between the male and the female groups. The results showed that survival probabilities decreased in both male and female groups after 10 months and leveled off after 30 months, with no statistically significant difference between the two groups (log-rank: p = 0.70).
Feature selection and model construction
To better understand whether systemic immune function reflects the progression and prognosis of AIS, we performed a group-wide study including 72 immunophenotypic indicators using flow cytometry on peripheral blood samples within 24 hours after the onset of AIS. We observed that among the 72 immunophenotypic indicators, we first excluded 31 indicators that were statistically different between the male and female groups to eliminate gender bias.pe Details were shown in Supplementary Table 2. Then, 22 of the 41 indicators were selected by univariate analysis of AIS versus controls (SupplementaryTable 3), and correlation analysis was performed on these 22 indicators, which showed that there was no strong correlation between them. Then, the least absolute shrinkage and selection operator (lasso) regression was applied to screen CD56highNK cells/μl, Tregs/μl, CD16+NK cells/μl, BM/μl, CTL (%), non-classical monocytes/μl and Monocytes/μl. These indicators were then applied to construct prognostic model (Figure 3).
Then, we performed univariate and multivariate Cox regression analysis to explore the immunophenotypic indicators associated with prognosis. And a forest plot was used to represent these indicators and their Hazard ratio (HR), 95% CI and p values between AIS and the control group, as detailed in Figure 4. Univariate analysis revealed that CTL (%) [HR: 2.03, 95% CI: 2.00-2.06, p< 0.001], Monocytes/μl [HR: 2.00, 95% CI: 1.90-2.10, p<0.001], Non-classical monocytes/μl [HR: 1.51, 95% CI: 1.20-1.82, p=0.015], CD56highNK cells/μl [HR:1.70, 95% CI:1.50-1.90, p=0.008] and CD16+NK cells/μl [HR:1.05, 95%CI:1.01-1.11, p=0.005] may contribute to decreasing the survival probability of AIS, while Tregs/μl [HR:0.96, 95%CI:0.94-0.99, p=0.003] and BM/μl [HR: 0.97, 95% CI: 0.95-0.99, p<0.001]were prognostic protective factors of AIS. On the other hand, multivariate Cox regression analysis showed that CTL (%) [HR: 1.18, 95% CI: 1.03-1.33, p=0.034], Monocytes/μl [HR: 1.13, 95% CI: 1.05-1.21, p=0.043], Non-classical monocytes/μl [HR: 1.09, 95% CI: 1.02-1.16, p=0.041] and CD56highNK cells/μl [HR: 1.13, 95% CI: 1.05-1.21, p<0.001] decreased the survival probability in the AIS group,, and Tregs/μl [HR:0.97, 95% CI: 0.95-0.99, p=0.004], BM/μl [HR:0.90, 95% CI: 0.85-0.95, p=0.023] and CD16+NK cells/μl [HR:0.93, 95% CI: 0.88-0.98, p=0.034] may have a protective effect on the prognostic likelihood of survival in patients with AIS. The differences between the groups of the above indicators were statistically significant (p<0.05).
Survival probability nomogram development and performance of the Cox model
Based on the above seven indicators, a multivariate Cox regression model was constructed, and the 7 indicators in the cox regression model were integrated to the nomogram to predict the survival rate of patients in the next 1, 2, and 3 years, respectively. For each AIS patient, the higher the total points, indicated the lower survival probability. For example, if the patient had a CTL (%) of 40, Tregs of 5μl, BM (%) of 10, monocytes of 600μl, non-classical monocytes of 80, CD56highNK cells of 25μl and CD16+NK cells of 30μl, then the corresponding points would be approximately 10, 10, 15, 10, 10, 40 and 50, respectively. The total score would be approximately 145, indicating an estimated survival probability of 48% for the next two years and 32% for the next three years for this case. More details can be found in Figure 5.
To assess the accuracy of predicting AIS adverse events, the AUC values of the univariate cox regression model ROC curves were calculated. The AUC of CD56highNK cells/ul was higher at 0.912 (0.884-0.954), while the AUC value of CD16+NK cells/μl was lower at less than 0.8 (0.72, 95% CI: 0.654-0.758). The accuracy of the other indicators in predicting AIS prognosis ranged from 0.820 to 0. 879. Detailed information on the AUC of each indicator and its 95% confidence interval was shown in Supplementary Figure 2. Based on the performance of these indicators in the univariate analysis, we combined seven indicators to construct a multivariate Cox model. The results showed that the AUC of the integrated Cox model was 0.805 (0.781–0.819), which could be used to predict the prognosis of AIS. To further validate the performance of the multivariate model in the prediction of AIS survival, we followed up 82 patients at 2 years and collected their peripheral blood immunophenotypic indicators as a test set for analysis. Notably, the multivariate Cox model achieved a high AUC of 0.961 (0.924-0.982) in the test set, indicating that the model is relatively stable and has a good predictive performance (Figure 6).
To further confirm the prognostic value of these immunophenotypic indicators, we employed stratified analysis and divide AIS patients into high and low subgroups based on the mean of each indicator. Notably, AIS patients with higher levels of Tregs/μl, BM/μl or CD16+NK cells/μl had higher survival probability than AIS patients with lower levels. In contrast, for AIS patients with lower levels of CD56highNK cells/μl, CTL (%), non-classical monocytes/μl or Monocytes/μl, survival probability may be higher than in the high-level group (Figure 7).
The heatmap showed the relationship between the seven immunophenotypic indicators used to construct the model and as the clinical characteristics, including gender, age (≤60 or >60 years), drinking and smoking status (yes, no and NA) and living status (alive, dead and NA) (Figure 8).