A total of 757 eligible CAPD patients with a mean follow-up time of 54.7 (standard deviation (SD), 33) months were included in the study (Figure 1). The median age of our cohort was 49 (interquartile range (IQR), 38 - 60) years, and the proportion of men was 55.1%. A total of 91 (12%) patients experienced stroke during a median follow-up time of 15 months and with a median occurrence age of 61.5 years and the counts of ischemic stroke and hemorrhagic stroke were 74 (83.1%) cases and 23 (25.8%) cases, with median ages of 64.5 and 55 years, respectively. The median incidence of stroke among our CAPD patients was 18.9 (IQR, 15.7 - 22.1) per 1000 person-years. Notably, patients had a higher incidence of stroke at CAPD initiation and 5 years and 10 years after CAPD (Supplemental Figure 1). One hundred fifty-three (20%) patients in our cohort experienced composite endpoints, and the proportion of composite endpoint events increased significantly in the stroke group compared to the non-stroke group (39.6% vs 17.6%, respectively; p-value < 0.001).
A few significant differences were observed at the initiation of CAPD between the stroke and non-stroke groups in our cohort. The median age of the stroke group was significantly older (62 vs 48 years; p-value < 0.001), and the stroke group had lower serum albumin (33.6 vs 35.7 g/l, p-value = 0.002), serum phosphorus (1.6 vs 1.7 mmol/l, p-value = 0.001), iPTH (167.7 vs 269.0 pg/ml, p-value = 0.001) and DBP (76.6 vs 84.6 mmHg, p-value < 0.001) levels. Furthermore, the prevalence of chronic heart disease (97.8% vs 25.8%, p-value <0.001) and diabetes (53.8% vs 24.3%, p-value < 0.001) was significantly higher in the stroke group. Interestingly, the prevalence of vitamin D supplementation was significantly lower in the stroke group than in the non-stroke group (53.8% vs 70.9%, respectively; p-value = 0.002). The results of the comparison between the stroke and non-stroke groups are shown in Table 1.
Difference in iPTH levels during follow-up between the stroke and non-stroke groups
Compared to the non-stroke group, the median values of the original, log-transformed and time-averaged iPTH levels during follow-up decreased significantly in the stroke group (p-values = 0.001, < 0.001 and 0.001, respectively). Furthermore, the median absolute difference in serum iPTH levels also decreased significantly in the stroke group (52.9 [34.4, 93.2] vs 66.2 [41.7, 106.5], p-value = 0.03), indicating that the serum iPTH level was significantly reduced in the stroke group compared with the non-stroke group during follow-up (Supplemental Table 1).
The nonlinear regression curves displayed markedly different trends in iPTH levels during follow-up in the stroke group and the non-stroke group. The iPTH levels gradually decreased in the stroke group, but increased in the non-stroke group as the number of months of dialysis increased (Figure 2).
Relationship between baseline iPTH and stroke
Our data showed a significantly skewed distribution of baseline serum iPTH levels, and the probability density distribution was markedly different between the non-stroke group and the stroke group, with a significant leftward shift in the peak value for the stroke group, indicating that the serum iPTH levels were significantly lower in the stroke group (Supplemental Figure 2a). Our nonlinear Cox regression analysis indicated a significant nonlinear correlation between baseline iPTH levels and the risk of stroke (p-value of the linear part = 0.2 and nonlinear part = 0.002). The curve of the relative stroke rate by baseline iPTH levels (referred to as 152 pg/ml) was J-shaped, suggesting that patients with low and markedly high levels of iPTH had a higher risk of stroke (Supplemental Figure 2b).
A Kaplan-Meier analysis of stroke among patients with different baseline serum iPTH levels showed a significant difference in the cumulative risk of stroke between the groups (log-rank test, p-value < 0.001), and patients with low baseline iPTH levels (≤150 pg/ml) had an increased cumulative risk of stroke. The pairwise comparison between groups revealed significant differences between the ≤150 group and the 150-300 group and the 300-600 group (p-values = 0.002 and <0.001, respectively), but significant differences were not observed between the ≤150 group and the > 600 group (p-value = 0.1, Figure 3).
Risk factors for stroke and composite endpoints
According to our univariate Cox regression analysis, increasing age and decreased DBP and iPTH levels combined with chronic heart disease, atrial fibrillation or diabetes, treatment with antiplatelet agents and not taking vitamin D supplements are common risk factors for stroke and composite endpoints (Table 2). Surprisingly, positive correlations were observed between the use of antiplatelet agents with the risk of stroke and composite endpoints, which we postulated might be a reverse epidemiological phenomenon, such as the treatment effect of statin on dialysis patients, or was attributed to the impact of the propensity treatment.
According to our data, a significant inverse correlation between serum iPTH levels and age (Kendall's rank correlation coefficient = -0.15, p-value < 0.001) was observed. Then, an interaction term of serum iPTH levels and age was included in the multivariate Cox models of stroke and composite endpoints (Table 3). In the stroke model (Table 3, model 1), our results revealed a significant effect of the interaction between serum iPTH levels and age; furthermore, the baseline serum iPTH levels showed a more significant association with stroke than age. In contrast to the stroke model, a significant effect of the interaction between serum iPTH levels and age was not observed on the composite model (Table 3, model 2). Based on the plot of the effects of the interaction between serum iPTH levels and age in the stroke model (Supplemental Figure 3), the preferred iPTH levels range from 150 and 300 pg/ml in younger patients (< 65 years) and 300 and 600 pg/ml in older patients (≤65 years).
Subgroup analysis for vitamin D supplementation
According to our multivariate Cox regression analysis, the consumption of a vitamin D supplement during follow-up was an independent protective factor both for stroke and the composite endpoints (model 1: HR, 0.42, 95% CI 0.24-0.74, p-value = 0.002; model 2: HR, 0.47, 95% CI 0.32-0.68, p-value < 0.001; Table 3). To further investigate the effects of vitamin D supplementation among different populations of CAPD patients, a subgroup analysis was performed. Regardless of the levels of serum calcium or phosphate, vitamin D supplementation was a significant protective factor for stroke. Interestingly, vitamin D supplementation was an independent predictive factor for stroke in male patients and older patients (HR 0.38, 95% CI 0.2-0.72, and HR 0.24, 95% CI 0.1-0.58, respectively). Additionally, vitamin D supplementation is associated with a reduced risk of stroke in patients with serum iPTH levels less than 600 pg/ml (Supplemental Figure 4).