Chronic Kidney Disease and the Risk of Stroke
We identified 83 eligible studies (63 cohort studies and 20 RCTs) including 2 253 741 participants with follow-up ranging from 0.25 months to 21 years (Figure 1 and Supplementary Data, Appendix Table S2 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1). In total, there were 30 392 all-cause stroke events including 21 633 which were not classified by pathological subtype (unspecified), 7498 ischaemic and 1625 haemorrhagic strokes. Characteristics of the included studies and randomized trials are described in Table 1 and Supplementary Data, Appendix Table S2 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1. Seventy-two studies (87%) reported unspecified stroke types, 23 studies (28%) reported ischaemic strokes and 18 studies (22%) reported haemorrhagic strokes (Table 1 and Figure 1). Twenty-seven (33%) of all studies were conducted in North America but studies which reported subtypes of stroke were more often conducted in Asia: 13 (57%) studies reporting ischaemic and 11 (61%) studies reporting haemorrhagic strokes. GFR was most commonly estimated using the MDRD formula (30 studies, 54% of studies reporting GFR) and albuminuria most commonly measured by ACR (17 studies, 46% of studies reporting ACR).
(Enlarge Image)
Figure 1.
Identification and inclusion of study reports of CKD and stroke risk.
Fifty-six (67%) studies (2 156 147 participants) assessed the association between GFR and stroke. We first analyzed data by stage of CKD. A GFR of <90 mL/min/1.73 m was associated with an increased risk of all-cause stroke by 39% (RR: 1.39, 95% CI: 1.31–1.47). The risk of all-cause stroke increased further with declining renal function. In participants with a GFR of 60–90 mL/min/1.73 m the risk of stroke was increased by 10% (RR: 1.10, 95% CI: 1.03–1.19), by 43% in participants with a GFR of 30–60 mL/min/1.73 m (RR: 1.43, 95% CI: 1.33–1.54) and by 70% in participants with an GFR of <30 mL/min/1.73 m (RR: 1.70, 95% CI: 1.47–1.96, test for difference, P < 0.001) (Figure 2). For every 10 mL/min/1.73 m decrease in GFR (relative to the reference group), the risk of having a stroke increased by 7% (RR: 1.07, 95% CI: 1.04–1.09) (Figure 3). In our multivariate-adjusted model, we observed 29% smaller risk estimates for stroke in larger studies (>20 000 participants) compared with smaller studies (<2500) (RR: 0.71, 95% CI: 0.55–0.91) and 27% smaller risk estimates for stroke in studies where effect estimates were adjusted compared with studies where estimates were unadjusted (RR: 0.73, 95% CI: 0.60–0.90) (Figure 1). Studies where participants underwent a heart procedure reported a 91% higher risk of stroke (RR: 1.91, 95% CI: 1.36–2.69). Risk of stroke did not vary among studies by any stroke characteristic including subtype, severity and whether incident or recurrent. Similarly, risk did not vary by the formula used to estimate GFR (Supplementary Data, Appendix Table S3 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1).
(Enlarge Image)
Figure 2.
Risk of all-cause stroke by GFR and albuminuria thresholds.
(Enlarge Image)
Figure 3.
RR of all-cause stroke by (i) decrease in GFR and (ii) increase in ACR from reference to comparison groups. Circles represent study-specific effect estimates and are proportional in size to the precision of effect estimates. GFR explains 67% and proteinuria 55% of total variation in risk of stroke after adjusting for other study/participant characteristics significantly associated with risk of stroke.
Thirty-seven (45%) studies (1 262 952 participants) assessed the association of albuminuria with stroke. Any degree of albuminuria increased the risk of all-cause stroke by 68% (RR: 1.68, 95% CI: 1.54–1.84). There was some evidence that the risk of all-cause strokes rose with increasing quantity of albuminuria, from 53% with microalbuminuria (RR: 1.53, 95% CI: 1.40–1.67) to 94% with macroalbuminuria (RR: 1.94, 95% CI 1.64–2.29, test for difference P = 0.06) (Figure 2). The risk of having a stroke increased by 10% (RR: 1.10, 95% CI: 1.01–1.20) for every 25 mg/mmol increase in ACR (Figure 3). In our multivariate-adjusted model, studies with a higher proportion of Asian participants reported a 75% increased risk of stroke (RR: 1.75; 95% CI: 1.07–2.86) compared with studies recruiting mainly white participants (Supplementary Data, Appendix Figure 2 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1). Albuminuria impacted risk of stroke differently according to age, with people aged 60–65 years experiencing a 45% smaller increased stroke risk than people aged <60 years (RR: 0.45, 95% CI: 0.28–0.71). The stroke subtype, severity and whether incident or recurrent had no effect on magnitude of risk estimates. Risk did not vary by the method used to quantify albuminuria (Supplementary Data, Appendix Table S3 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1).
Eight studies (12%) examined the association between GFR with albuminuria and stroke, but we could use data from only six studies because two studies used a reference range of GFR other than >60 mL/min/1.73 m or a reference range of albuminuria other than none. Risk of stroke in participants with a GFR <60 mL/min/1.73 m and any albuminuria (RR: 2.18, 95% CI: 1.68–2.84) was additive, approximately equal to the sum of the risk of stroke in participants with GFR <60 mL/min/1.73 m (RR: 1.51, 95% CI: 1.41–1.61) and the risk of stroke among participants with any albuminuria (RR: 1.68, 95% CI: 1.54–1.84). In our multivariate-adjusted regression models, we tested for interaction between GFR and albuminuria and found none detectable.
Excluding randomized controlled trials where attrition bias was judged unclear ( seven studies) and cohort studies where estimates of effect were unadjusted for at least age, gender and smoking (47 studies) did not alter the association we observed for the association between either GFR or albuminuria and the risk of all, unspecified, ischaemic or haemorrhagic stroke (Supplementary Data, Appendix Table S4 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1).
Results
We identified 83 eligible studies (63 cohort studies and 20 RCTs) including 2 253 741 participants with follow-up ranging from 0.25 months to 21 years (Figure 1 and Supplementary Data, Appendix Table S2 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1). In total, there were 30 392 all-cause stroke events including 21 633 which were not classified by pathological subtype (unspecified), 7498 ischaemic and 1625 haemorrhagic strokes. Characteristics of the included studies and randomized trials are described in Table 1 and Supplementary Data, Appendix Table S2 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1. Seventy-two studies (87%) reported unspecified stroke types, 23 studies (28%) reported ischaemic strokes and 18 studies (22%) reported haemorrhagic strokes (Table 1 and Figure 1). Twenty-seven (33%) of all studies were conducted in North America but studies which reported subtypes of stroke were more often conducted in Asia: 13 (57%) studies reporting ischaemic and 11 (61%) studies reporting haemorrhagic strokes. GFR was most commonly estimated using the MDRD formula (30 studies, 54% of studies reporting GFR) and albuminuria most commonly measured by ACR (17 studies, 46% of studies reporting ACR).
(Enlarge Image)
Figure 1.
Identification and inclusion of study reports of CKD and stroke risk.
GFR and Stroke
Fifty-six (67%) studies (2 156 147 participants) assessed the association between GFR and stroke. We first analyzed data by stage of CKD. A GFR of <90 mL/min/1.73 m was associated with an increased risk of all-cause stroke by 39% (RR: 1.39, 95% CI: 1.31–1.47). The risk of all-cause stroke increased further with declining renal function. In participants with a GFR of 60–90 mL/min/1.73 m the risk of stroke was increased by 10% (RR: 1.10, 95% CI: 1.03–1.19), by 43% in participants with a GFR of 30–60 mL/min/1.73 m (RR: 1.43, 95% CI: 1.33–1.54) and by 70% in participants with an GFR of <30 mL/min/1.73 m (RR: 1.70, 95% CI: 1.47–1.96, test for difference, P < 0.001) (Figure 2). For every 10 mL/min/1.73 m decrease in GFR (relative to the reference group), the risk of having a stroke increased by 7% (RR: 1.07, 95% CI: 1.04–1.09) (Figure 3). In our multivariate-adjusted model, we observed 29% smaller risk estimates for stroke in larger studies (>20 000 participants) compared with smaller studies (<2500) (RR: 0.71, 95% CI: 0.55–0.91) and 27% smaller risk estimates for stroke in studies where effect estimates were adjusted compared with studies where estimates were unadjusted (RR: 0.73, 95% CI: 0.60–0.90) (Figure 1). Studies where participants underwent a heart procedure reported a 91% higher risk of stroke (RR: 1.91, 95% CI: 1.36–2.69). Risk of stroke did not vary among studies by any stroke characteristic including subtype, severity and whether incident or recurrent. Similarly, risk did not vary by the formula used to estimate GFR (Supplementary Data, Appendix Table S3 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1).
(Enlarge Image)
Figure 2.
Risk of all-cause stroke by GFR and albuminuria thresholds.
(Enlarge Image)
Figure 3.
RR of all-cause stroke by (i) decrease in GFR and (ii) increase in ACR from reference to comparison groups. Circles represent study-specific effect estimates and are proportional in size to the precision of effect estimates. GFR explains 67% and proteinuria 55% of total variation in risk of stroke after adjusting for other study/participant characteristics significantly associated with risk of stroke.
Albuminuria and Stroke
Thirty-seven (45%) studies (1 262 952 participants) assessed the association of albuminuria with stroke. Any degree of albuminuria increased the risk of all-cause stroke by 68% (RR: 1.68, 95% CI: 1.54–1.84). There was some evidence that the risk of all-cause strokes rose with increasing quantity of albuminuria, from 53% with microalbuminuria (RR: 1.53, 95% CI: 1.40–1.67) to 94% with macroalbuminuria (RR: 1.94, 95% CI 1.64–2.29, test for difference P = 0.06) (Figure 2). The risk of having a stroke increased by 10% (RR: 1.10, 95% CI: 1.01–1.20) for every 25 mg/mmol increase in ACR (Figure 3). In our multivariate-adjusted model, studies with a higher proportion of Asian participants reported a 75% increased risk of stroke (RR: 1.75; 95% CI: 1.07–2.86) compared with studies recruiting mainly white participants (Supplementary Data, Appendix Figure 2 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1). Albuminuria impacted risk of stroke differently according to age, with people aged 60–65 years experiencing a 45% smaller increased stroke risk than people aged <60 years (RR: 0.45, 95% CI: 0.28–0.71). The stroke subtype, severity and whether incident or recurrent had no effect on magnitude of risk estimates. Risk did not vary by the method used to quantify albuminuria (Supplementary Data, Appendix Table S3 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1).
GFR With Albuminuria and Stroke
Eight studies (12%) examined the association between GFR with albuminuria and stroke, but we could use data from only six studies because two studies used a reference range of GFR other than >60 mL/min/1.73 m or a reference range of albuminuria other than none. Risk of stroke in participants with a GFR <60 mL/min/1.73 m and any albuminuria (RR: 2.18, 95% CI: 1.68–2.84) was additive, approximately equal to the sum of the risk of stroke in participants with GFR <60 mL/min/1.73 m (RR: 1.51, 95% CI: 1.41–1.61) and the risk of stroke among participants with any albuminuria (RR: 1.68, 95% CI: 1.54–1.84). In our multivariate-adjusted regression models, we tested for interaction between GFR and albuminuria and found none detectable.
Sensitivity Analyses
Excluding randomized controlled trials where attrition bias was judged unclear ( seven studies) and cohort studies where estimates of effect were unadjusted for at least age, gender and smoking (47 studies) did not alter the association we observed for the association between either GFR or albuminuria and the risk of all, unspecified, ischaemic or haemorrhagic stroke (Supplementary Data, Appendix Table S4 http://ndt.oxfordjournals.org/content/30/7/1162/suppl/DC1).
Source...