In the prospective national stroke registry, we found SDNN, a sensitive index of HRV, was negatively associated with both the inflammatory markers and stroke outcomes in patients with acute ischemic stroke or TIA. Two biomarkers of systematic inflammation, hs-CRP and IL-6, partially mediated the association between SDNN and stroke outcomes. In contrast, RMSSD, another index of HRV, was not associated with inflammatory markers and stroke outcomes.
The relationship between autonomic nervous function and stroke has been evaluated by several studies2, 22, 23. In line with our findings, a previous study proposed that ANS dysfunction was associated with an increased risk of stroke recurrence and death22. Meanwhile, SDNN was also found to be associated with risk of stroke and all-cause death2, 23. The atherosclerosis risk in communities’ study with longer follow-up showed lower HRV (measured by time-domain and frequency-domain) was associated with a higher risk of incident stroke, and SDNN and RMSSD were both the good predictive factors for stroke risk3. However, the present study found that only SDNN, rather than RMSSD, was associated with stroke outcomes. The potential explanation of positive relationship for SDNN, rather than RMSSD, might be that RMSSD reflects parasympathetic nerve, which changes are transient and variable, whereas SDNN reflects the excitability of both sympathetic and parasympathetic nerves and the sympathetic excitation period is longer after ischemic stroke. Difference in outcome (recurrent stroke vs incident stroke), time of follow-up (one year vs. longer up to 10 years) and HRV measurement method (time-domain vs. frequency-domain) may also account for the different findings between our study and previous studies3, 22–24.
Many studies have confirmed a continuous cross-talk between both sympathetic and parasympathetic branches of the autonomic nervous system and inflammatory response in different clinical scenarios9, 25, 26. The study about vagal nerve pointed that HRV could predict inflammation-mediated death in patients with pancreatic cancer, which suggested that inflammation may be a mediator of some diseases27. A meta-analysis on HRV and inflammation showed the importance of the vagus nerve in regulating and controlling the inflammatory response, and there was an overall negative relationship between HRV and markers of inflammation9. Previous studies, including the Heart and Soul Study, showed that SDNN, rather than RMSSD, was negatively related to hs-CRP and IL-6, which was consistent with our findings25, 26. Furthermore, three previous studies have shown that changes in HRV often occurred before abnormal changes in inflammatory marker indicators, which added evidences that HRV might influence stroke outcome by regulating inflammation11, 12, 27. Our study further added evidence supporting that the association between HRV, especially SDNN, and stroke outcome was partially mediated by inflammatory marker.
The possible reason why IL-6 and hs-CRP, rather than YKL-40 and LP-PLA2-A, showed mediation effects might be that IL-6 and hs-CRP are markers of systemic inflammation. Whereas, LP-PLA2 was related to the plaque inflammation/endothelial dysfunction, and YKL-40 was an inflammatory biomarker involved in regulating glial activation and neuroinflammation28–30. Moreover, we also found that although hs-CRP and IL-6 mediated the relationship between HRV and stroke recurrence, the mediating percentage of IL-6 was larger than that of hs-CRP. The cytokine of hs-CRP was considered a downstream biomarker in the IL-1, IL-6, hs-CRP pathway; whereas, IL-6 was an upstream signaling cytokine, which has become a major target for immune regulation and thrombosis protection in atherosclerosis31. Therefore, as the indicator of inflammation, IL-6 may have higher sensitivity and accuracy. Nevertheless, it should be acknowledged that the mediating percentage of IL-6 was still somewhat small, suggesting that inflammation might only be a partial mediating factor. There might be other pathways or mechanisms involved in mediating the association between autonomic nervous system and stroke outcome which warranted further research.
Inflammation is closely related to atherosclerotic disease, and inflammatory regulation can prevent clinical complications of atherosclerosis32. Previous study demonstrated that inflammation was associated with the residual risk of stroke recurrence for ischemic stroke with adherence to guideline-based secondary stroke prevention13. Inflammatory response was controlled, in part, by the neural circuitry of the autonomic nervous system5. The nervous system, via the anti-inflammatory pathway, could inhibit cytokine synthesis and protect against inflammatory cytokine-mediated diseases33. Our study demonstrated that inflammatory markers partially mediated the relationship between HRV and stroke outcomes. This implicates that reducing inflammation by regulating HRV might improve stroke prognosis. However, this needs more rational research in future.
This study has several limitations. First, the mediating proportion of our study is relatively low, so inflammatory markers may not be the most important mediating and influencing factors. Second, although we used the inverse weighted probability method to deal with the missing value, selection bias caused by missing data still might exist. Third, since Cox proportional hazard model was rarely used in mediation analysis, the evidence on these models is limited, leaving room for unknown bias. Fourth, HRV and related blood sample data are both collected during hospitalization. We can't accurately judge the successive relationship between them in time. Furthermore, inflammatory markers and HRV might be affected by the stress function in acute ischemic stroke patients, reflecting only be temporarily high or low rather than steady-state data. Fifth, hs-CRP and IL-6 were not the most upstream inflammatory markers and couldn't represent all inflammatory response. Sixth, we did not verify the validity of the prediction model in other external cohorts. Therefore, the accurate mediation effect of inflammation for the association between ANS and the stroke outcomes might need more investigation in large-scale prospective studies with more rigorous protocols or sophisticated instruments.