Chronic Kidney Disease and the Risk of Stroke
Chronic Kidney Disease and the Risk of Stroke
In this meta-analysis of 83 studies, which included data for >30 000 strokes in 2 253 741 participants, we have demonstrated a linear relationship between decreasing GFR and the risk of stroke, and between increasing albuminuria and the risk of stroke. Specifically, we observed an average 7% increase in the RR of stroke associated with every 10 mL/min/1.73 m reduction in GFR, and an average 10% increase in the RR of stroke associated with every 25 mg/mmol increase in ACR. Findings were consistent when GFR was analyzed by stage of CKD and albuminuria analyzed by category of albuminuria. Our analyses also provide compelling evidence about the thresholds of GFR and albuminuria at which the risk of having a stroke starts to increase with a GFR below 90 mL/min/1.73 m and microalbuminuria each independently associated with an increased stroke risk.
Our observations were consistent across different subtypes of stroke, in men and women, and among studies where participants had varying levels of risk factors for vascular disease (diabetes, hypertension and smoking). Study size, adjustment for potential confounders, the clinical setting of the study and the race of study participants all affected the size of the estimated risk of stroke. In particular, we observed a 91% increased risk of stroke in studies where participants with reduced GFR were undergoing a heart procedure. For albuminuria, risk of stroke was 75% greater in studies which recruited mainly South-East Asian participants compared with studies where most participants were white, an observation which has major public health implications. Data were lacking for patients with end stage kidney disease requiring renal replacement therapy.
Based on our data, we propose that the revised Kidney Disease: Improving Global Outcomes (KDIGO) classification of CKD may be a useful tool for stratifying the risk of stroke in the general population though estimates of absolute risk of stroke require adjustment for race and clinical setting (Figure 4).
(Enlarge Image)
Figure 4.
Ten year absolute risk of all-cause stroke (fatal and non-fatal).
To our knowledge, this study forms the largest and most comprehensive review of stroke in relation to kidney function and is the first meta-analysis to consider CKD in terms of both GFR and albuminuria. In identifying potential studies for inclusion, we performed a comprehensive literature search and included a large number of studies, so that estimates of effect were generally precise. We included only RCTs and cohort studies and consequently, selection bias, recall bias and reverse causality are unlikely to have affected our results. We also used validated tools to assess the quality of included studies and undertook sensitivity analyses to determine how robust our effect estimates were to varying levels of quality of discreet study methodology. Our analyses also present the most comprehensive exploration of heterogeneity among studies examining the associations of GFR and albuminuria with the risk of stroke.
Our study does have some potential limitations. First, most included studies were observational and we were limited to examining heterogeneity based on data published for known cardiovascular risk factors. Many studies did not report data for known confounders of stroke risk, particularly the use of treatments for complications of CKD (including erythropoietin for anaemia and angiotensin II enzyme inhibitors for people with albuminuria) or the reduction of cardiovascular risk (including aspirin and statins) which may affect the risk of ischaemic and haemorrhagic stroke differently. Residual confounding may exist within our results and explain why we observed no difference in the associations between CKD and the risk of ischaemic or haemorrhagic strokes. Second, the diagnosis of CKD requires the presence of kidney damage for ≥3 months. GFR or albuminuria was frequently only measured once meaning that some patients with acute kidney injury or non-persistent albuminuria may have been misclassified as having CKD and that our estimates of effect may be subject to regression dilution bias. Estimates of GFR may be subject to misclassification bias. The MDRD equation underestimates GFR and the Cockcroft-Gault equation overestimates GFR at >60 mL/min/1.73 m in healthy individuals, and both equations overestimate GFR at >60 mL/min/1.73 m in people with reduced muscle mass (who are more likely to be unwell and may be at higher risk of stroke). We also estimated average levels of GFR and albuminuria within category ranges based on previously described methods which assume a near-normal distribution of GFR or albuminuria within each defined CKD stage or category of albuminuria, an assumption we could not test without individual patient data. Finally, the lack of confirmation of type of stroke (ischaemic versus haemorrhagic) in many studies limited our ability to explore stroke subtype in detail.
The magnitude of increased risk of stroke we observed in study participants with a GFR < 60 mL/min/1.73 m is similar to that seen in a previous meta-analysis of 21 prospective studies, but the GFR threshold at which we began to see risk of stroke increase (<90 mL/min/1.73 m) suggests an earlier stage of CKD than previously thought. A previous review that included 10 studies identified albuminuria as being associated with an increased risk of stroke but did not observe an increasing risk with increasing quantities of albuminuria or any variation in the effect of albuminuria by race.
Our findings for stroke are also broadly consistent with the effects of GFR and albuminuria on the risk of other cardiovascular events. The magnitudes of risk of stroke associated with GFR and albuminuria are similar to size of risk of myocardial infarction and cardiovascular mortality, which are increased by 33 and 57% when GFR is <60 mL/min/1.73 m and by 48 and 63% with microalbuminuria. The threshold GFR at which risk begins to increase is <60 mL/min/1.73 m for myocardial infarction and <75 mL/min/1.73 m for cardiovascular death. For albuminuria, increased risk of myocardial infarction and cardiovascular death are also seen with micro albuminuria. Linear dose–response relationships exist between GFR, quantity of albuminuria and risk of myocardial infarction and cardiovascular death. GFR and albuminuria independently increase the risk of myocardial infarction and cardiovascular mortality without interaction. Considering effect modifiers of GFR, although age and ethnicity modify the effect of GFR on cardiovascular death, we did not observe this for stroke. Patients with lower GFRs undergoing heart procedures are at increased risk of myocardial infarction, heart failure and death compared with patients with higher GFRs, and our observation of increased stroke risk is consistent with these established associations. For albuminuria, no good data on factors that modify the effect of albuminuria on cardiovascular events exist.
There are several possible explanations for finding a near-doubled risk of stroke in studies with mainly Asian participants who had albuminuria, most of which assume that increased exposure to albuminuria increases the risk of stroke. Hypertension occurs at a younger age and may cause more profound end-organ damage in Asians than in whites, and no study that we included adjusted for the duration of hypertension. Diabetes occurs a decade earlier, at lower body mass index and is more frequently associated with albuminuria and worse glycaemic control in Asians than whites. Although we adjusted for differences in the prevalence of diabetes, we could not adjust for either the duration of diabetes or glycaemic control. Asians with albuminuria are also more likely to have a lower GFR than whites with albuminuria and not all studies adjusted for the effect of GFR on the association between albuminuria and stroke.
Plausible pathological mechanisms provide further support for GFR and/or albuminuria having a role in causing stroke in addition to the strength, consistency and biological gradient of effect that we have demonstrated. Endothelial dysfunction is common in CKD in which uraemic toxins, insulin resistance, vascular calcification, dyslipidaemia, anaemia and renin–angiotensin activation are proposed to cause chronic inflammation, oxidative stress and promote atherogenesis and arteriosclerosis. To date, data from RCTs demonstrating a reduction in the risk of stroke coincident with slowing the rate of GFR decline or reducing albuminuria is lacking with only a few conflicting reports for other cardiovascular endpoints. Synthesis of eight RCTs failed to show any reduction in the risk of cardiovascular mortality amongst participants randomized to treatments known to slow decline in GFR. For albuminuria, one RCT reported that reducing albuminuria by 50% reduced the risk of a cardiovascular event.
For researchers, the challenge is to generate further evidence that preventing progressive CKD reduces the risk of having a stroke. Specifically, the effects of reducing albuminuria in South-East Asians, slowing the rate of change in GFR and how dynamic changes in the quantity of albuminuria affect stroke risk should be examined. The public health implications of our data are also considerable and the potential for stroke prevention is substantial. Our data suggest that each year up to 4% of all strokes (31 800 in the USA) may be attributable to having a GFR <90 mL/min/1.73 m, having any degree of albuminuria may account for 6% (47 770) of all strokes, and that as many as 10 000 (1.2%) strokes could be prevented if people with microalbuminuria received an angiotensin-converting enzyme inhibitor.
Discussion
In this meta-analysis of 83 studies, which included data for >30 000 strokes in 2 253 741 participants, we have demonstrated a linear relationship between decreasing GFR and the risk of stroke, and between increasing albuminuria and the risk of stroke. Specifically, we observed an average 7% increase in the RR of stroke associated with every 10 mL/min/1.73 m reduction in GFR, and an average 10% increase in the RR of stroke associated with every 25 mg/mmol increase in ACR. Findings were consistent when GFR was analyzed by stage of CKD and albuminuria analyzed by category of albuminuria. Our analyses also provide compelling evidence about the thresholds of GFR and albuminuria at which the risk of having a stroke starts to increase with a GFR below 90 mL/min/1.73 m and microalbuminuria each independently associated with an increased stroke risk.
Our observations were consistent across different subtypes of stroke, in men and women, and among studies where participants had varying levels of risk factors for vascular disease (diabetes, hypertension and smoking). Study size, adjustment for potential confounders, the clinical setting of the study and the race of study participants all affected the size of the estimated risk of stroke. In particular, we observed a 91% increased risk of stroke in studies where participants with reduced GFR were undergoing a heart procedure. For albuminuria, risk of stroke was 75% greater in studies which recruited mainly South-East Asian participants compared with studies where most participants were white, an observation which has major public health implications. Data were lacking for patients with end stage kidney disease requiring renal replacement therapy.
Based on our data, we propose that the revised Kidney Disease: Improving Global Outcomes (KDIGO) classification of CKD may be a useful tool for stratifying the risk of stroke in the general population though estimates of absolute risk of stroke require adjustment for race and clinical setting (Figure 4).
(Enlarge Image)
Figure 4.
Ten year absolute risk of all-cause stroke (fatal and non-fatal).
To our knowledge, this study forms the largest and most comprehensive review of stroke in relation to kidney function and is the first meta-analysis to consider CKD in terms of both GFR and albuminuria. In identifying potential studies for inclusion, we performed a comprehensive literature search and included a large number of studies, so that estimates of effect were generally precise. We included only RCTs and cohort studies and consequently, selection bias, recall bias and reverse causality are unlikely to have affected our results. We also used validated tools to assess the quality of included studies and undertook sensitivity analyses to determine how robust our effect estimates were to varying levels of quality of discreet study methodology. Our analyses also present the most comprehensive exploration of heterogeneity among studies examining the associations of GFR and albuminuria with the risk of stroke.
Our study does have some potential limitations. First, most included studies were observational and we were limited to examining heterogeneity based on data published for known cardiovascular risk factors. Many studies did not report data for known confounders of stroke risk, particularly the use of treatments for complications of CKD (including erythropoietin for anaemia and angiotensin II enzyme inhibitors for people with albuminuria) or the reduction of cardiovascular risk (including aspirin and statins) which may affect the risk of ischaemic and haemorrhagic stroke differently. Residual confounding may exist within our results and explain why we observed no difference in the associations between CKD and the risk of ischaemic or haemorrhagic strokes. Second, the diagnosis of CKD requires the presence of kidney damage for ≥3 months. GFR or albuminuria was frequently only measured once meaning that some patients with acute kidney injury or non-persistent albuminuria may have been misclassified as having CKD and that our estimates of effect may be subject to regression dilution bias. Estimates of GFR may be subject to misclassification bias. The MDRD equation underestimates GFR and the Cockcroft-Gault equation overestimates GFR at >60 mL/min/1.73 m in healthy individuals, and both equations overestimate GFR at >60 mL/min/1.73 m in people with reduced muscle mass (who are more likely to be unwell and may be at higher risk of stroke). We also estimated average levels of GFR and albuminuria within category ranges based on previously described methods which assume a near-normal distribution of GFR or albuminuria within each defined CKD stage or category of albuminuria, an assumption we could not test without individual patient data. Finally, the lack of confirmation of type of stroke (ischaemic versus haemorrhagic) in many studies limited our ability to explore stroke subtype in detail.
The magnitude of increased risk of stroke we observed in study participants with a GFR < 60 mL/min/1.73 m is similar to that seen in a previous meta-analysis of 21 prospective studies, but the GFR threshold at which we began to see risk of stroke increase (<90 mL/min/1.73 m) suggests an earlier stage of CKD than previously thought. A previous review that included 10 studies identified albuminuria as being associated with an increased risk of stroke but did not observe an increasing risk with increasing quantities of albuminuria or any variation in the effect of albuminuria by race.
Our findings for stroke are also broadly consistent with the effects of GFR and albuminuria on the risk of other cardiovascular events. The magnitudes of risk of stroke associated with GFR and albuminuria are similar to size of risk of myocardial infarction and cardiovascular mortality, which are increased by 33 and 57% when GFR is <60 mL/min/1.73 m and by 48 and 63% with microalbuminuria. The threshold GFR at which risk begins to increase is <60 mL/min/1.73 m for myocardial infarction and <75 mL/min/1.73 m for cardiovascular death. For albuminuria, increased risk of myocardial infarction and cardiovascular death are also seen with micro albuminuria. Linear dose–response relationships exist between GFR, quantity of albuminuria and risk of myocardial infarction and cardiovascular death. GFR and albuminuria independently increase the risk of myocardial infarction and cardiovascular mortality without interaction. Considering effect modifiers of GFR, although age and ethnicity modify the effect of GFR on cardiovascular death, we did not observe this for stroke. Patients with lower GFRs undergoing heart procedures are at increased risk of myocardial infarction, heart failure and death compared with patients with higher GFRs, and our observation of increased stroke risk is consistent with these established associations. For albuminuria, no good data on factors that modify the effect of albuminuria on cardiovascular events exist.
There are several possible explanations for finding a near-doubled risk of stroke in studies with mainly Asian participants who had albuminuria, most of which assume that increased exposure to albuminuria increases the risk of stroke. Hypertension occurs at a younger age and may cause more profound end-organ damage in Asians than in whites, and no study that we included adjusted for the duration of hypertension. Diabetes occurs a decade earlier, at lower body mass index and is more frequently associated with albuminuria and worse glycaemic control in Asians than whites. Although we adjusted for differences in the prevalence of diabetes, we could not adjust for either the duration of diabetes or glycaemic control. Asians with albuminuria are also more likely to have a lower GFR than whites with albuminuria and not all studies adjusted for the effect of GFR on the association between albuminuria and stroke.
Plausible pathological mechanisms provide further support for GFR and/or albuminuria having a role in causing stroke in addition to the strength, consistency and biological gradient of effect that we have demonstrated. Endothelial dysfunction is common in CKD in which uraemic toxins, insulin resistance, vascular calcification, dyslipidaemia, anaemia and renin–angiotensin activation are proposed to cause chronic inflammation, oxidative stress and promote atherogenesis and arteriosclerosis. To date, data from RCTs demonstrating a reduction in the risk of stroke coincident with slowing the rate of GFR decline or reducing albuminuria is lacking with only a few conflicting reports for other cardiovascular endpoints. Synthesis of eight RCTs failed to show any reduction in the risk of cardiovascular mortality amongst participants randomized to treatments known to slow decline in GFR. For albuminuria, one RCT reported that reducing albuminuria by 50% reduced the risk of a cardiovascular event.
For researchers, the challenge is to generate further evidence that preventing progressive CKD reduces the risk of having a stroke. Specifically, the effects of reducing albuminuria in South-East Asians, slowing the rate of change in GFR and how dynamic changes in the quantity of albuminuria affect stroke risk should be examined. The public health implications of our data are also considerable and the potential for stroke prevention is substantial. Our data suggest that each year up to 4% of all strokes (31 800 in the USA) may be attributable to having a GFR <90 mL/min/1.73 m, having any degree of albuminuria may account for 6% (47 770) of all strokes, and that as many as 10 000 (1.2%) strokes could be prevented if people with microalbuminuria received an angiotensin-converting enzyme inhibitor.
Source...