Abstract
Background:
Few studies have investigated the association between serum albumin levels and the risk of stroke subtypes among the general Japanese population.
Methods and Results:
In this study, 5,071 men and 7,969 women aged 40–74 years, initially free from stroke, coronary artery disease, and kidney and hepatic failure, and residing in 4 Japanese communities completed a baseline risk factor survey between 1985 and 1994. During the 24-year follow-up, 528 men and 553 women experienced stroke. In the entire study cohort, multivariable hazard ratios (HRs) and 95% confidence intervals (CIs) of total stroke, ischemic stroke, and intracerebral hemorrhage for the lowest vs. highest quartiles of serum albumin were 1.45 (1.18–1.77), 1.52 (1.17–1.97), and 1.57 (1.04–2.37), respectively. In men, multivariable HRs (95% CIs) for total stroke, ischemic stroke, and intracerebral hemorrhage in the lowest vs. highest serum albumin quartile were 1.44 (1.07–1.92), 1.48 (1.03–2.11) and 1.71 (0.92–3.18), respectively, whereas in women they were 1.50 (1.13–1.99), 1.63 (1.11–2.39), and 1.56 (0.89–2.74), respectively. Similar inverse associations were observed for each of the ischemic stroke subtypes, but not for subarachnoid hemorrhage.
Conclusions:
Low serum albumin levels were associated with an increased risk of total stroke, ischemic stroke, ischemic stroke subtypes, and intracerebral hemorrhage.
Albumin, the most abundant protein in human blood plasma, has essential physiologic effects in maintaining health. Albumin is a multifunctional protein that modulates the colloid osmotic pressure between the blood vessels and tissues, and binds and transports various endogenous molecules (e.g., free fatty acids, hormones, bilirubin, and metal ions) and some medications. The synthesis of albumin is adversely affected by poor nutritional status and chronic inflammation.1
Albumin is considered an antioxidant because of its ligand-binding and free radical-scavenging properties.2
Furthermore, albumin has been reported to exert anticoagulant actions and inhibitory effects on platelet function.3
Numerous Western and Asian cohort studies have reported that low serum albumin concentrations are independent risk factors of all-cause and cardiovascular mortality in the general population.4–9
Several Western cohort studies further reported an inverse association between serum albumin concentrations and the risk of total and ischemic strokes;10–12
in one of these studies, which examined the risks of ischemic stroke subtypes, low serum albumin concentrations were found to be associated with an increased risk of cardioembolic and cryptogenic ischemic stroke.12
Stroke is a dominant subtype of cardiovascular disease in most Asian countries.13
However, few Asian cohort studies have investigated associations between serum albumin concentrations and the risk of stroke and stroke subtypes.
The aim of the present study was to examine associations between serum albumin concentrations and the risk of stroke and stroke subtypes among middle-aged Japanese men and women. We hypothesized that low serum albumin concentrations were associated with an increased risk of stroke and stroke subtypes, independent of serum total cholesterol concentrations and other traditional cardiovascular risk factors.
Methods
Study Population
This study is part of the Circulatory Risk in Communities Study (CIRCS), an ongoing dynamic community cohort study of cardiovascular disease in the general Japanese population ongoing since 1963.14,15
The study population was comprised of 5,263 men and 8,081 women community residents aged 40–74 years who participated in annual health checkups. The participants were enrolled from 4 communities: Ikawa (a rural community in Akita Prefecture in northwestern Japan), Minami-Takayasu (a suburb in Osaka Prefecture in mid-western Japan), Noichi (a rural community in Kochi Prefecture in western Japan), and Kyowa (a rural community in Ibaraki Prefecture in mid-eastern Japan). Baseline surveys were conducted in these 4 communities in 1985–1990, 1985–1994, 1985–1990 and 1985–1991, respectively. Individuals who had a history of stroke, coronary artery disease, hepatic failure, kidney failure, or were undergoing hemodialysis at baseline (193 men, 111 women) were excluded. This left 5,071 men and 7,969 women available for the present analysis. Informed consent was obtained from community representatives because this study was a secondary use of existing data from the cardiovascular disease prevention program in Japanese communities. Ethics approval was obtained from the ethics committees of the Osaka Center for Cancer and Cardiovascular Disease Prevention (Reference no. 29-Ethics-2) and Osaka University (Reference no. 14285-6).
Follow-up and Ascertainment of Cases
Participants were followed up to determine incident stroke, and its subtypes, occurring by the end of 2010 for Noichi, 2014 for Kyowa, 2016 for Yao, and 2017 for Ikawa. Follow-up was terminated at the first incident stroke, exit from the community, or death; 1,020 (8%) participants moved out from the communities and 4,633 (36%) died. The median follow-up period was 24 years.
Details regarding endpoint determination have been described in previous reports of CIRCS.14,15
Stroke surveillance was performed during the whole follow-up period. Information for candidate cases of stroke was ascertained from multiple sources, such as death certificates, national insurance claims, annual household questionnaires, annual cardiovascular risk surveys, and reports by either local physicians, public health nurses, or health volunteers. The diagnosis of stroke was further confirmed: all living patients with a suspected stroke were telephoned, visited, or invited to take part in a face-to-face survey during their annual health checkups. In addition, medical records were obtained from local clinics and hospitals. In the case of death, medical histories were obtained from families and/or attending physicians, and relevant medical records were reviewed. Stroke was defined as a focal neurological disorder with rapid in onset that persisted at least 24 h or until death. Stroke subtypes, including intracerebral and subarachnoid hemorrhage, ischemic stroke (lacunar infarction, large artery occlusive infarction, embolic infarction and unclassified infarction), were classified on the basis of computed tomography (CT) and magnetic resonance imaging (MRI) findings. For diagnosed cases of stroke without brain imaging, stroke subtypes were classified as ischemic stroke, intracerebral hemorrhage, subarachnoid hemorrhage, and unclassified stroke according to clinical criteria.16
Lacunar infarction was diagnosed as 1 or multiple infarctions involving focal, small, and deep areas based on the presence of lacunar syndrome and/or brain imaging, without cerebral cortical or cerebellar impairment. Large artery occlusive infarction was diagnosed as infarction involving the cortical artery regions in the cerebrum and cerebellum (cortex and subcortical areas) based on the presence of cortical signs and/or brain imaging. Embolic infarction was defined as cerebral infarction caused by emboli from extracranial sources. When an embolic source was present in the medical record and brain imaging supported the diagnosis, those infarctions were considered as embolic infarctions. Unclassified infarction included those cases of ischemic stroke that failed to meet the criteria for lacunar infarction, large artery occlusive infarction, or embolic infarction because of a lack of brain imaging to confirm the diagnosis. CT or MRI findings were available for 94% of stroke cases in this analysis. The final diagnoses were made by a panel of 2–4 experienced physician epidemiologists who were blinded to the data from the risk factor survey using the same diagnostic criteria for stroke.
Baseline Examination
Blood was drawn from seated subjects into plain, siliconized glass tubes, and the serum was separated within 30 min. Serum albumin was measured using the bromocresol green method. Serum total cholesterol was measured using the direct Lieberman-Burchard method from 1984 to August 31, 1986 and the enzymatic method from September 1, 1986 to 1994. Serum triglycerides were measured using the fluorometric method from 1984 to August 31, 1986, the enzymatic method from September 1, 1986 to July 22, 1993, and the enzymatic method for free glycerol from July 23, 1993 to 1994. Serum glucose was measured using the cupric-neocuproine method from 1984 to August 31, 1986, the hexokinase method from September 1, 1986 to July 22, 1993, and the glucokinase method from July 23, 1993 to 1994. Serum glucose concentrations (mmol/L) measured by the cupric-neocuproine method were adjusted using the following linear regression formula: serum glucose concentrations (mg/dL) × 0.0474 + 0.541. All measurements were performed at the Osaka Medical Central for Cancer and Cardiovascular Disease, an international member of the US National Cholesterol Reference Method Laboratory Network (CRMLN).17,18
During health checkups, subjects’ height (in stockinged feet) and weight (in light clothing) were measured. Body mass index (BMI) was calculated as weight (kg) divided by the height squared (m2). Trained observers interviewed participants to determine smoking status, the number of cigarettes smoked per day, usual weekly intake of alcohol (evaluated using the unit “go”, a traditional Japanese unit of volume corresponding to 23 g ethanol), and medication use. Menopausal status was ascertained in women, with no menstruation for more than 6 months defined as a postmenopausal status. Systolic and diastolic blood pressure (SBP/DBP) in the right arm were measured by trained physicians using standard mercury sphygmomanometers and unified epidemiological methods.19
Hypertension was defined as SBP ≥140 mmHg and/or DBP ≥90 mmHg and/or the use of antihypertensive medication. Diabetes was defined as fasting glucose levels ≥7.0 mmol/L and/or non-fasting glucose levels ≥11.1 mmol/L and/or the use of medication for diabetes.
Statistical Analyses
Analyses of covariance (ANCOVAs) were used to compare differences in sex-specific, age- and community-adjusted mean values or the prevalence of baseline characteristics according to serum albumin quartiles. Cox proportional hazards models were used to calculate hazard ratios (HRs) and 95% confidence interval (CIs) of stroke and its subtypes for each quartile and 1 standard deviation (SD) decrease in serum albumin (0.3 g/dL in all participants, 0.3 g/dL in men and 0.2 g/dL in women). Median values for each serum albumin quartile were using to test the trend of associations with the risk of stroke and its subtypes.
The first HR model was adjusted for age, sex, and community, whereas the second was adjusted for age, sex, community, and serum total cholesterol (mmol/L). The full multivariable model was further adjusted for sex-specific quartiles of BMI (kg/m2), cigarette smoking status (never, former, and current [1–19 or 20 cigarette/day]), alcohol intake status (never, former, and current [<23, 23–45, ≥46 g ethanol/day]), SBP (mmHg), antihypertensive medication use (no or yes), sex-specific quartiles of serum glutamic oxaloacetic transaminase (GOT) and glutamic pyruvic transaminase (GPT), sex-specific quartiles of serum triglycerides (mmol/L), atrial fibrillation (no or yes), diabetes (no or yes), and menopausal status (pre- or post-menopausal) in women. We evaluated the effect of the interaction between sex and serum albumin concentrations on the risk of total stroke and its subtypes using a cross-product term of sex (0 or 1) and serum albumin concentrations (continuous) for this model.
All statistical analyses were performed using SAS System for Windows version 9.4 (SAS Institute, Cary, NC, USA) and 2-tailed P<0.05 was considered significant.
Results
Table 1
lists sex-specific, age- and community-adjusted mean values or the prevalence of traditional cardiovascular risk factors at baseline according to serum albumin quartiles. A significant inverse association between age and serum albumin concentrations was observed in both sexes, and this association was stronger in men than in women. Serum albumin concentrations were positively associated with SBP and DBP, the use of antihypertensive medication, and serum total cholesterol and triglycerides concentrations in both sexes, as well as BMI in men, but inversely associated with the prevalence of current smoking and ethanol intake in men. Serum albumin concentrations were positively associated with the prevalence of diabetes and postmenopausal status in women. With regard to biomarkers of liver disease, serum albumin concentrations were positively associated with serum GOT and GPT in women, but tended to be inversely associated with both these biomarkers in men.
Table 1.
Baseline Characteristics of Participants According to Serum Albumin Quartiles
| |
Serum albumin quartile |
P value for
difference |
| Q1 (low) |
Q2 |
Q3 |
Q4 (high) |
| Men |
| No. at risk |
988 |
1,328 |
1,488 |
1,267 |
|
| Serum albumin (g/dL) |
| Range |
2.6–4.2 |
4.3–4.4 |
4.5–4.6 |
4.7–5.6 |
|
| Median |
4.1 |
4.4 |
4.5 |
4.8 |
|
| Age (years) |
59.6±0.3 |
56.5±0.2 |
53.5±0.2 |
50.1±0.2 |
<0.001 |
| BMI (kg/m2) |
22.4±0.1 |
23.0±0.1 |
23.3±0.1 |
23.5±0.1 |
<0.001 |
| SBP (mmHg) |
130.5±0.6 |
131.7±0.5 |
134.5±0.5 |
137.1±0.5 |
<0.001 |
| DBP (mmHg) |
79.1±0.4 |
80.5±0.3 |
83.4±0.3 |
85.0±0.3 |
<0.001 |
| Antihypertensive medication use (%) |
11.4 |
11.4 |
13.8 |
18.0 |
<0.001 |
| Serum TC (mmol/L) |
4.55±0.03 |
4.76±0.02 |
4.97±0.02 |
5.21±0.02 |
<0.001 |
| Serum TG (mmol/L) |
1.52±0.04 |
1.70±0.04 |
1.84±0.04 |
2.08±0.04 |
<0.001 |
| Serum GOT (IU/L) |
33.3±0.7 |
28.7±0.6 |
29.0±0.6 |
30.8±0.6 |
0.01 |
| Serum GPT (IU/L) |
34.3±1.0 |
30.8±0.8 |
32.2±0.8 |
33.8±0.9 |
0.03 |
| Diabetes (%) |
8.4 |
7.8 |
8.1 |
10.0 |
0.21 |
| Atrial fibrillation (%) |
1.5 |
1.0 |
0.8 |
1.1 |
0.47 |
| Current smoker (%) |
64.4 |
64.8 |
59.3 |
56.3 |
0.004 |
| Ethanol intake (g/day) |
32.1±0.9 |
29.5±0.8 |
28.4±0.7 |
28.7±0.8 |
0.01 |
| Women |
| No. at risk |
1,601 |
2,418 |
2,431 |
1,519 |
|
| Serum albumin (g/dL) |
| Range |
3.4–4.2 |
4.3–4.4 |
4.5–4.6 |
4.7–5.3 |
|
| Median |
4.2 |
4.4 |
4.5 |
4.8 |
|
| Age (years) |
55.3±0.2 |
54.0±0.2 |
53.8±0.2 |
53.5±0.2 |
<0.001 |
| BMI (kg/m2) |
23.3±0.1 |
23.3±0.1 |
23.5±0.1 |
23.5±0.1 |
0.09 |
| SBP (mmHg) |
127.6±0.4 |
130.3±0.4 |
132.3±0.4 |
135.4±0.5 |
<0.001 |
| DBP (mmHg) |
76.1±0.3 |
78.2±0.2 |
79.5±0.2 |
81.5±0.3 |
<0.001 |
| Antihypertensive medication use (%) |
10.9 |
13.6 |
14.7 |
19.2 |
<0.001 |
| Serum TC (mmol/L) |
4.87±0.02 |
5.09±0.02 |
5.31±0.02 |
5.54±0.02 |
<0.001 |
| Serum TG (mmol/L) |
1.38±0.03 |
1.47±0.02 |
1.56±0.02 |
1.74±0.03 |
<0.001 |
| Serum GOT (IU/L) |
23.5±0.3 |
23.5±0.2 |
24.1±0.2 |
25.2±0.3 |
<0.001 |
| Serum GPT (IU/L) |
22.6±0.5 |
22.2±0.4 |
22.9±0.4 |
24.3±0.5 |
0.007 |
| Diabetes (%) |
3.6 |
3.7 |
5.5 |
4.9 |
0.008 |
| Atrial fibrillation (%) |
0.4 |
0.5 |
0.2 |
0.2 |
0.19 |
| Current smoker (%) |
8.0 |
7.5 |
8.5 |
10.0 |
0.40 |
| Ethanol intake (g/day) |
1.6±0.2 |
1.5±0.1 |
1.3±0.1 |
1.4±0.2 |
0.68 |
| Postmenopausal (%) |
57.8 |
61.4 |
64.6 |
67.3 |
<0.001 |
Unless indicated otherwise, data are presented as the mean±SEM or proportions, adjusted for age and community. BMI, body mass index; DBP, diastolic blood pressure; GOT, glutamic oxaloacetic transaminase; GPT, glutamic pyruvic transaminase; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides.
During the median 24-year follow-up, totaling 288,770 person-years, there were 1,081 cases of incident stroke (528 in men, 553 in women). The observed stroke subtypes included 719 ischemic strokes (383 in men, 336 in women), 229 intracerebral hemorrhages (102 in men, 127 in women), 116 subarachnoid hemorrhages (34 in men, 82 in women), and 17 unclassified strokes (9 in men, 8 in women). Ischemic stroke subtypes included 325 lacunar infarctions (182 in men, 143 in women), 72 large artery occlusive infarctions (33 in men, 39 in women), 120 embolic infarctions (61 in men, 59 in women), and 202 unclassified infarctions (107 in men, 95 in women).
Table 2
lists the multivariable HRs for total stroke, ischemic stroke, and intracerebral and subarachnoid hemorrhages in all participants. Low serum albumin concentrations tended to be associated with increased age-, sex- and community-adjusted risks of total stroke, ischemic stroke, and intracerebral hemorrhage, but not subarachnoid hemorrhage. After adjustment for serum total cholesterol, these associations did not change materially. However, these associations were strengthened and became statistically significant after further adjustment for SBP and the use of antihypertensive medication. After adjustment for other conventional cardiovascular risk factors, these associations remained significant. The multivariable HRs of total stroke, ischemic stroke, and intracerebral hemorrhage for the lowest vs. highest quartile of serum albumin in all participants were 1.45 (95% CI 1.18–1.77; Ptrend
<0.001), 1.52 (95% CI 1.17–1.97; Ptrend=0.002), and 1.57 (95% CI 1.04–2.37; Ptrend=0.03), respectively. The multivariable HRs (95% CIs) of total stroke, ischemic stroke, and intracerebral hemorrhage for a 1-SD decrease in serum albumin concentrations were 1.14 (1.08–1.22), 1.16 (1.08–1.26) and 1.17 (1.02–1.33), respectively. Sex-specific HRs for total stroke and its subtypes are listed in
Table 3. Inverse associations between serum albumin concentrations and the risk of total and ischemic strokes were similarly observed in men and women, with no sex interaction (Pinteraction=0.33 for total stroke and Pinteraction=0.40 for ischemic stroke). Low serum albumin concentrations trended to be associated with increased risk of intracerebral hemorrhage in both men and women, and no association was observed for risk of subarachnoid hemorrhage in either sex.
Table 2.
HRs and 95% CIs of Total Stroke and Its Subtypes According to Serum Albumin Quartiles in All Participants
| |
Serum albumin quartile |
Ptrend |
1-SD decreaseA |
| Q1 (low) |
Q2 |
Q3 |
Q4 (high) |
| Serum albumin range (g/dL) |
2.6–4.2 |
4.3–4.4 |
4.5–4.6 |
4.7–5.6 |
|
|
| No. at risk |
2,589 |
3,746 |
3,919 |
2,786 |
|
|
| Person-years |
53,084 |
82,818 |
88,380 |
64,488 |
|
|
| Total stroke |
| No. events |
261 |
314 |
309 |
197 |
|
|
Age-, sex-, community-adjusted
HR (95% CI) |
1.16 (0.96–1.40) |
1.04 (0.87–1.24) |
1.03 (0.86–1.23) |
1.00 |
0.12 |
1.07 (1.01–1.13) |
| Multivariable HR (95% CI) |
| Model 1 |
1.17 (0.96–1.43) |
1.05 (0.87–1.26) |
1.03 (0.86–1.23) |
1.00 |
0.10 |
1.08 (1.01–1.14) |
| Model 2 |
1.37 (1.12–1.67) |
1.15 (0.96–1.39) |
1.10 (0.92–1.32) |
1.00 |
0.002 |
1.13 (1.06–1.20) |
| Model 3 |
1.45 (1.18–1.77) |
1.19 (0.99–1.44) |
1.13 (0.94–1.35) |
1.00 |
<0.001 |
1.14 (1.08–1.22) |
| Ischemic stroke |
| No. events |
165 |
228 |
213 |
113 |
|
|
Age-, sex-, community-adjusted
HR (95% CI) |
1.18 (0.92–1.51) |
1.26 (1.00–1.58) |
1.20 (0.96–1.51) |
1.00 |
0.22 |
1.08 (1.00–1.16) |
| Multivariable HR (95% CI) |
| Model 1 |
1.24 (0.96–1.60) |
1.30 (1.03–1.64) |
1.22 (0.97–1.54) |
1.00 |
0.11 |
1.10 (1.02–1.19) |
| Model 2 |
1.44 (1.12–1.87) |
1.43 (1.13–1.80) |
1.30 (1.04–1.64) |
1.00 |
0.005 |
1.15 (1.07–1.24) |
| Model 3 |
1.52 (1.17–1.97) |
1.45 (1.15–1.84) |
1.34 (1.06–1.69) |
1.00 |
0.002 |
1.16 (1.08–1.26) |
| Intracerebral hemorrhage |
| No. events |
66 |
54 |
61 |
48 |
|
|
Age-, sex-, community-adjusted
HR (95% CI) |
1.30 (0.88–1.91) |
0.77 (0.52–1.13) |
0.85 (0.58–1.25) |
1.00 |
0.17 |
1.10 (0.97–1.25) |
| Multivariable HR (95% CI) |
| Model 1 |
1.20 (0.81–1.80) |
0.73 (0.49–1.09) |
0.83 (0.57–1.22) |
1.00 |
0.32 |
1.08 (0.95–1.23) |
| Model 2 |
1.40 (0.94–2.10) |
0.81 (0.54–1.21) |
0.89 (0.60–1.30) |
1.00 |
0.09 |
1.13 (0.99–1.29) |
| Model 3 |
1.57 (1.04–2.37) |
0.88 (0.59–1.32) |
0.93 (0.63–1.36) |
1.00 |
0.03 |
1.17 (1.02–1.33) |
| Subarachnoid hemorrhage |
| No. events |
24 |
29 |
30 |
33 |
|
|
Age-, sex-, community-adjusted
HR (95% CI) |
0.77 (0.45–1.32) |
0.62 (0.38–1.03) |
0.62 (0.38–1.02) |
1.00 |
0.28 |
0.93 (0.78–1.10) |
| Multivariable HR (95% CI) |
| Model 1 |
0.75 (0.43–1.31) |
0.61 (0.37–1.02) |
0.62 (0.37–1.01) |
1.00 |
0.25 |
0.92 (0.77–1.09) |
| Model 2 |
0.89 (0.50–1.56) |
0.69 (0.41–1.15) |
0.66 (0.40–1.09) |
1.00 |
0.57 |
0.97 (0.81–1.15) |
| Model 3 |
0.94 (0.53–1.67) |
0.72 (0.43–1.22) |
0.70 (0.42–1.15) |
1.00 |
0.73 |
0.98 (0.82–1.17) |
AThe 1-SD decrease in serum albumin was 0.3 g/dL. Model 1 was further adjusted for serum TC; Model 2 was further adjusted for SBP and the use of antihypertensive medication; and Model 3 was further adjusted for BMI, smoking status, alcohol intake status, serum triglycerides, GOT, and GPT concentrations, atrial fibrillation, and diabetes. CIs, confidence Intervals; HRs, hazard ratios. Other abbreviations as in Table 1.
Table 3.
Sex-Specific HRs
A and 95% CIs for Total Stroke and Its Subtypes According to Serum Albumin Quartiles
| |
Serum albumin quartile |
Ptrend |
1-SD decreaseB |
| Q1 (low) |
Q2 |
Q3 |
Q4 (high) |
| Men |
| Serum albumin range (g/dL) |
2.6–4.2 |
4.3–4.4 |
4.5–4.6 |
4.7–5.6 |
|
|
| No. at risk |
988 |
1,328 |
1,488 |
1,267 |
|
|
| Person-years |
17,767 |
27,019 |
31,741 |
28,457 |
|
|
| Total stroke |
| No. events |
126 |
149 |
146 |
107 |
|
|
Age- and community-adjusted
HR (95% CI) |
1.14 (0.86–1.50) |
1.03 (0.80–1.33) |
0.99 (0.77–1.28) |
1.00 |
0.34 |
1.09 (0.98–1.21) |
| Multivariable HR (95% CI) |
| Model 1 |
1.19 (0.90–1.59) |
1.06 (0.82–1.38) |
1.01 (0.79–1.31) |
1.00 |
0.21 |
1.12 (1.00–1.25) |
| Model 2 |
1.40 (1.05–1.86) |
1.18 (0.91–1.53) |
1.09 (0.84–1.40) |
1.00 |
0.02 |
1.18 (1.06–1.32) |
| Model 3 |
1.44 (1.07–1.92) |
1.18 (0.91–1.54) |
1.11 (0.86–1.43) |
1.00 |
0.01 |
1.19 (1.07–1.33) |
| Ischemic stroke |
| No. events |
86 |
124 |
107 |
66 |
|
|
Age- and community-adjusted
HR (95% CI) |
1.13 (0.80–1.59) |
1.29 (0.95–1.75) |
1.13 (0.83–1.54) |
1.00 |
0.46 |
1.09 (0.97–1.24) |
| Multivariable HR (95% CI) |
| Model 1 |
1.21 (0.85–1.72) |
1.35 (0.99–1.84) |
1.16 (0.85–1.58) |
1.00 |
0.26 |
1.13 (0.99–1.29) |
| Model 2 |
1.39 (0.98–1.99) |
1.47 (1.08–2.02) |
1.24 (0.90–1.69) |
1.00 |
0.05 |
1.19 (1.05–1.35) |
| Model 3 |
1.48 (1.03–2.11) |
1.49 (1.09–2.04) |
1.30 (0.95–1.78) |
1.00 |
0.03 |
1.21 (1.06–1.37) |
| Intracerebral hemorrhage |
| No. events |
29 |
18 |
30 |
25 |
|
|
Age- and community-adjusted
HR (95% CI) |
1.41 (0.78–2.53) |
0.63 (0.34–1.18) |
0.96 (0.56–1.65) |
1.00 |
0.33 |
1.13 (0.88–1.43) |
| Multivariable HR (95% CI) |
| Model 1 |
1.43 (0.78–2.63) |
0.64 (0.34–1.20) |
0.97 (0.56–1.67) |
1.00 |
0.31 |
1.14 (0.88–1.46) |
| Model 2 |
1.74 (0.94–3.20) |
0.73 (0.39–1.39) |
1.05 (0.61–1.80) |
1.00 |
0.10 |
1.22 (0.96–1.56) |
| Model 3 |
1.71 (0.92–3.18) |
0.72 (0.38–1.37) |
1.01 (0.59–1.75) |
1.00 |
0.12 |
1.23 (0.96–1.58) |
| Subarachnoid hemorrhage |
| No. events |
7 |
6 |
8 |
13 |
|
|
Age- and community-adjusted
HR (95% CI) |
0.93 (0.34–2.53) |
0.51 (0.19–1.36) |
0.56 (0.23–1.36) |
1.00 |
0.61 |
0.93 (0.61–1.42) |
| Multivariable HR (95% CI) |
| Model 1 |
0.83 (0.29–2.35) |
0.47 (0.17–1.30) |
0.53 (0.22–1.31) |
1.00 |
0.50 |
0.89 (0.57–1.38) |
| Model 2 |
1.02 (0.36–2.91) |
0.56 (0.20–1.57) |
0.58 (0.23–1.42) |
1.00 |
0.79 |
0.98 (0.64–1.52) |
| Model 3 |
1.02 (0.35–3.02) |
0.60 (0.21–1.67) |
0.61 (0.24–1.52) |
1.00 |
0.80 |
0.98 (0.60–1.53) |
| Women |
| Serum albumin range (g/dL) |
3.4–4.2 |
4.3–4.4 |
4.5–4.6 |
4.7–5.3 |
|
|
| No. at risk |
1,601 |
2,418 |
2,431 |
1,519 |
|
|
| Person-years |
35,318 |
55,799 |
56,638 |
36,031 |
|
|
| Total stroke |
| No. events |
135 |
165 |
163 |
90 |
|
|
Age- and community-adjusted
HR (95% CI) |
1.27 (0.97–1.67) |
1.11 (0.86–1.44) |
1.10 (0.85–1.42) |
1.00 |
0.08 |
1.08 (1.00–1.16) |
| Multivariable HR (95% CI) |
| Model 1 |
1.23 (0.93–1.63) |
1.09 (0.84–1.41) |
1.08 (0.84–1.40) |
1.00 |
0.14 |
1.07 (0.99–1.15) |
| Model 2 |
1.44 (1.09–1.91) |
1.20 (0.92–1.56) |
1.16 (0.89–1.50) |
1.00 |
0.01 |
1.11 (1.03–1.20) |
| Model 3 |
1.50 (1.13–1.99) |
1.22 (0.94–1.59) |
1.18 (0.91–1.53) |
1.00 |
0.005 |
1.12 (1.04–1.20) |
| Ischemic stroke |
| No. events |
79 |
104 |
106 |
47 |
|
|
Age- and community-adjusted
HR (95% CI) |
1.34 (0.93–1.94) |
1.30 (0.92–1.84) |
1.34 (0.95–1.89) |
1.00 |
0.16 |
1.08 (0.99–1.19) |
| Multivariable HR (95% CI) |
| Model 1 |
1.36 (0.93–1.99) |
1.31 (0.93–1.87) |
1.34 (0.95–1.90) |
1.00 |
0.15 |
1.09 (0.99–1.20) |
| Model 2 |
1.61 (1.10–2.36) |
1.46 (1.03–2.08) |
1.44 (1.02–2.03) |
1.00 |
0.02 |
1.14 (1.04–1.25) |
| Model 3 |
1.63 (1.11–2.39) |
1.45 (1.02–2.07) |
1.44 (1.02–2.04) |
1.00 |
0.02 |
1.13 (1.03–1.24) |
| Intracerebral hemorrhage |
| No. events |
37 |
36 |
31 |
23 |
|
|
Age- and community-adjusted
HR (95% CI) |
1.41 (0.83–2.40) |
0.97 (0.57–1.64) |
0.83 (0.48–1.42) |
1.00 |
0.13 |
1.13 (0.97–1.31) |
| Multivariable HR (95% CI) |
| Model 1 |
1.20 (0.69–2.08) |
0.88 (0.51–1.49) |
0.79 (0.46–1.35) |
1.00 |
0.37 |
1.07 (0.92–1.25) |
| Model 2 |
1.36 (0.78–2.37) |
0.95 (0.56–1.63) |
0.83 (0.48–1.43) |
1.00 |
0.19 |
1.11 (0.95–1.30) |
| Model 3 |
1.56 (0.89–2.74) |
1.03 (0.60–1.78) |
0.88 (0.51–1.52) |
1.00 |
0.08 |
1.15 (0.99–1.35) |
| Subarachnoid hemorrhage |
| No. events |
17 |
23 |
22 |
20 |
|
|
Age- and community-adjusted
HR (95% CI) |
0.84 (0.43–1.62) |
0.74 (0.41–1.35) |
0.69 (0.38–1.27) |
1.00 |
0.59 |
0.95 (0.79–1.15) |
| Multivariable HR (95% CI) |
| Model 1 |
0.81 (0.41–1.60) |
0.73 (0.39–1.34) |
0.69 (0.37–1.26) |
1.00 |
0.55 |
0.94 (0.78–1.14) |
| Model 2 |
0.96 (0.48–1.90) |
0.81 (0.44–1.49) |
0.74 (0.40–1.36) |
1.00 |
0.88 |
0.99 (0.82–1.20) |
| Model 3 |
0.99 (0.50–1.98) |
0.82 (0.44–1.52) |
0.77 (0.42–1.42) |
1.00 |
0.95 |
0.99 (0.82–1.21) |
AThe P values for the sex interaction were 0.33 for total stroke, 0.40 for ischemic stroke, 0.39 for intracerebral hemorrhage and 0.58 for subarachnoid hemorrhage. BThe 1-SD decrease in serum albumin was 0.3 g/dL in men and 0.2 g/dL in women. Model 1 was further adjusted for serum TC; Model 2 was further adjusted for SBP and the use of antihypertensive medication; and Model 3 was further adjusted for BMI, smoking status, alcohol intake status, serum triglycerides, GOT, and GPT concentrations, atrial fibrillation, and diabetes. Abbreviations as in Tables 1,2.
When we examined ischemic stroke subtypes, low serum albumin concentrations were associated with an increased risk of lacunar infarction in the total study cohort; the multivariable HR for the lowest vs. highest quartile of serum albumin was 1.82 (95% CI 1.24–2.67; Ptrend=0.002), whereas that for a 1-SD decrease in serum albumin concentrations was 1.18 (95% CI 1.06–1.33). In addition, a 1-SD decrease in serum albumin concentrations was associated with an increased risk of large artery occlusive and embolic infarctions, with multivariable HRs (95% CIs) of 1.26 (1.00–1.60) and 1.24 (1.03–1.48), respectively (Supplementary Table 1). Inverse associations with serum albumin concentrations were similarly observed for lacunar infarction, primarily in men, and for large artery occlusive and embolic infarctions in women (Supplementary Table 2).
Discussion
The present prospective community-based study of 5,071 Japanese men and 7,969 women community residents aged 40–74 years found that low serum albumin concentrations were associated with increased risks of total stroke, ischemic stroke, ischemic stroke subtypes, and intracerebral hemorrhage, but not subarachnoid hemorrhage, after adjustment for serum total cholesterol, SBP, the use of antihypertensive medication, and other conventional cardiovascular risk factors.
The inverse associations between serum albumin concentrations and the of total and ischemic strokes in the present study are consistent with results reported in previous studies. The First National Health and Nutrition Examination Survey (NHANES I) of 4,157 US Whites and 740 Blacks aged 45–74 years with 9–16 years of follow-up reported that serum albumin concentrations were inversely associated with the risk of incident total and non-hemorrhagic strokes; with multivariable HRs (95% CIs) of total and non-hemorrhagic strokes for the highest (>4.4 g/dL) vs. lowest (<4.2 g/dL) tertiles of serum albumin being 0.59 (0.37–0.93) and 0.58 (0.36–0.93), respectively, in participants aged 45–64 years and 0.74 (0.57–0.94) and 0.72 (0.55–0.94), respectively, in those aged 65–74 years.10
The British Regional Heart Study (BRHS) of 7,690 British men aged 40–59 years with a 16.8-year follow-up reported a similar inverse association, with a multivariable HR (95% CI) of incident total stroke for the highest (≥4.7 g/dL) vs. lowest (<4.3 g/dL) quartiles of serum albumin of 0.63 (0.44–0.89).11
The Northern Manhattan Study of 2,986 American men and women aged ≥40 years with a 12-year follow-up period investigated the associations of serum albumin concentrations with risks of incident ischemic stroke subtypes.12
That study found that low serum albumin was associated with increased risks of cardioembolic and cryptogenic ischemic strokes, with the multivariable HRs (95% CIs) for the lowest (2.7–4.2 g/dL) vs. highest (4.6–5.5 g/dL) tertiles of serum albumin being 0.60 (0.24–1.52) for lacunar infarction (n=65), 1.36 (0.73–2.55) for large artery occlusive infarction (n=35), 1.92 (1.10–3.34) for cardioembolic infarction (n=92), and 2.59 (1.21–5.53) for cryptogenic ischemic stroke (n=55).12
The Northern Manhattan Study did not provide data on intracerebral and subarachnoid hemorrhages. Cryptogenic ischemic stroke is considered a subtype of ischemic stroke; it does not have a well-defined etiology, with possible mechanisms including occult paroxysmal atrial fibrillation and other atrial cardiopathies, paradoxical embolism through a patent foramen ovale, and substenotic atherosclerosis, among others.20
However, this stroke subtype was not considered in the present study because of the difficulties associated with the consistent detection and diagnosis of cryptogenic ischemic stroke.
Several potential mechanisms underlying the inverse relationship between serum albumin concentrations and the risk of ischemic stroke have been proposed. First, low serum albumin concentrations are indicators of moderate to severe malnutrition. Second, low serum albumin concentrations are an indication of inflammation: inflammatory activation of macrophages and other immune system cells produces more cytokines (e.g., interleukin-1, interleukin-6 and tumor necrosis factor-α), causing protein synthesis in the liver to switch from albumin to other acute-phase proteins.21
Third, albumin has antioxidant effects by binding copper, iron, and other cationic ligands, inhibiting the generation of reactive oxygen species.22,23
Albumin also has indirect antioxidant effects by binding bilirubin, with the albumin-bound bilirubin protecting against oxidative damage caused by low-density lipoprotein cholesterol.24
Fourth, albumin plays an important role in anticoagulation by binding antithrombin4
and its inhibitory effects on platelet aggregation.25,26
To the best of our best knowledge, this study is the first to find an excess risk of intracerebral hemorrhage associated with low serum albumin concentrations (<4.3 g/dL) after adjustment for traditional cardiovascular risk factors. The plausibility of this excess risk has been reported elsewhere. Low serum albumin concentrations were associated with low animal protein intake.27
Both the Nurses’ Health Study cohort study of 85,764 American women aged 34–59 years28
and the Hisayama Study of 2,400 Japanese men and women aged 40–79 years29
showed that low animal protein intake was associated with an increased risk of intracerebral hemorrhage, with the multivariable HRs (95% CI) for the highest vs. lowest quantiles of animal protein intake of 0.32 (95% CI 0.10–1.00; Ptrend=0.04) and 0.47 (95% CI 0.23–0.96; Ptrend=0.03), respectively.
Serum total cholesterol, which, like albumin, is synthesized in the liver, is another useful indicator of nutritional status. Accordingly, serum albumin concentrations were positively associated with serum total cholesterol in the baseline surveys, but the association of serum albumin concentrations with the risk of total stroke or its subtypes did not change materially after adjustment for serum total cholesterol concentrations. In addition, serum albumin concentrations were positively associated with baseline SBP and DBP, as well as the use of antihypertensive medication, which is consistent with the findings of a previous study.30
After further adjustment for SBP and the use of antihypertensive medication, the inverse associations between serum albumin concentrations and the risk of total stroke or its subtypes strengthened and became statistically significant.
The strength of the present study is its large-scale prospective cohort design with a median follow-up of 24 years, so that sufficient cases of stroke were collected to investigate sex-specific associations between serum albumin concentrations and the risk of stroke and its subtypes. However, the study does have some limitations. First, the single measurement of serum albumin at the time of the baseline survey could bias the association towards nil, such that the real associations would be greater than those we found. Second, the number of cases of large artery occlusive and embolic infarctions was relatively small, so that further follow-up is needed to confirm our results regarding ischemic stroke subtypes. Third, other inflammatory markers (e.g., C-reactive protein and leukocyte counts) were not considered because of limited data.
In conclusion, low serum albumin concentrations were associated with an increased risk of total stroke, ischemic stroke, ischemic stroke subtypes, and intracerebral hemorrhage, but not subarachnoid hemorrhage.
Acknowledgments
The authors thank their colleagues from Osaka University Center of Medical Data Science, Advanced Clinical Epidemiology Investigator’s Research Project for providing insights into and their expertise for this study.
Grants
This study was supported by Grant-in-Aid for Co-operative Research (A) [grant number 04304036], Scientific Research B [grant numbers 08457125, 10470103 and 12470092], and Challenging Research (Exploratory) [grant number 17K19810] from the Japan Society for the Promotion of Science.
Disclosures
None declared.
IRB Information
This study was approved by the ethics committees of the Osaka Center for Cancer and Cardiovascular Disease Prevention (Reference no. 29-Ethics-2) and Osaka University (Reference no. 14285-6).
Appendix
The CIRCS Investigators are:
• Takeo Okada, Yuji Shimizu, Yasuhiko Kubota, Shinichi Sato, Mina Hayama-Terada, and Masahiko Kiyama (Osaka Center for Cancer and Cardiovascular Disease Prevention)
• Hironori Imano, Renzhe Cui, Isao Muraki, Akihiko Kitamura, Hiroshige Jinnouchi, Mizuki Sata, and Hiroyasu Iso (Osaka University)
• Kazumasa Yamagishi, Mitsumasa Umesawa, and Tomoko Sankai (University of Tsukuba).
• Koutatsu Maruyama (Ehime University)
• Ai Ikeda and Takeshi Tanigawa (Juntendo University)
• Masanori Nagao and Tetsuya Ohira (Fukushima Medical University)
Supplementary Files
Please find supplementary file(s);
http://dx.doi.org/10.1253/circj.CJ-20-0384
References
- 1.
Don BR, Kaysen G. Serum albumin: Relationship to inflammation and nutrition. Semin Dial 2004; 17: 432–437.
- 2.
Roche M, Rondeau P, Singh NR, Tarnus E, Bourdon E. The antioxidant properties of serum albumin. FEBS Lett 2008; 582: 1783–1787.
- 3.
Paar M, Rossmann C, Nusshold C, Wagner T, Schlagenhauf A, Leschnik B, et al. Anticoagulant action of low, physiologic, and high albumin levels in whole blood. PLoS One 2017; 12: e0182997.
- 4.
Danesh J, Collins R, Appleby P, Peto R. Association of fibrinogen, C-reactive protein, albumin, or leukocyte count with coronary heart disease: Meta-analyses of prospective studies. JAMA 1998; 279: 1477–1482.
- 5.
Phillips A, Shaper AG, Whincup PH. Association between serum albumin and mortality from cardiovascular disease, cancer, and other causes. Lancet 1989; 2: 1434–1436.
- 6.
Corti MC, Guralnik JM, Salive ME, Sorkin JD. Serum albumin level and physical disability as predictors of mortality in older persons. JAMA 1994; 272: 1036–1042.
- 7.
Djoussé L, Rothman KJ, Cupples LA, Levy D, Ellison RC. Serum albumin and risk of myocardial infarction and all-cause mortality in the Framingham Offspring Study. Circulation 2002; 106: 2919–2924.
- 8.
Okamura T, Hayakawa T, Kadowaki T, Kita Y, Okayama A, Elliott P, et al. A combination of serum low albumin and above-average cholesterol level was associated with excess mortality. J Clin Epidemiol 2004; 57: 1188–1195.
- 9.
Umeki Y, Adachi H, Enomoto M, Fukami A, Nakamura S, Nohara Y, et al. Serum albumin and cerebro-cardiovascular mortality during a 15-year study in a community-based cohort in Tanushimaru, a cohort of the Seven Countries Study. Intern Med 2016; 55: 2917–2925.
- 10.
Gillum RF, Ingram DD, Makuc DM. Relation between serum albumin concentration and stroke incidence and death: The NHANES I Epidemiologic Follow-up Study. Am J Epidemiol 1994; 140: 876–888.
- 11.
Shaper AG, Wannamethee SG, Whincup PH. Serum albumin and risk of stroke, coronary heart disease, and mortality: The role of cigarette smoking. J Clin Epidemiol 2004; 57: 195–202.
- 12.
Xu WH, Dong C, Rundek T, Elkind MS, Sacco RL. Serum albumin levels are associated with cardiac embolic and cryptogenic ischemic strokes: Northern Manhattan Study. Stroke 2014; 45: 973–978.
- 13.
Ohira T, Iso H. Cardiovascular disease epidemiology in Asia: An overview. Circ J 2013; 77: 1646–1652.
- 14.
Imano H, Kitamura A, Sato S, Kiyama M, Ohira T, Yamagishi K, et al. Trends for blood pressure and its contribution to stroke incidence in the middle-aged Japanese population: The Circulatory Risk in Communities Study (CIRCS). Stroke 2009; 40: 1571–1577.
- 15.
Yamagishi K, Muraki I, Kubota Y, Hayama-Terada M, Imano H, Cui R, et al. The Circulatory Risk in Communities Study (CIRCS): A long-term epidemiological study for lifestyle-related disease among Japanese men and women living in communities. J Epidemiol 2019; 29: 83–91.
- 16.
Iso H, Rexrode K, Hennekens CH, Manson JE. Application of computer tomography-oriented criteria for stroke subsubtype classification in a prospective study. Ann Epidemiol 2000; 10: 81–87.
- 17.
Nakamura M, Iso H, Kitamura A, Imano H, Kiyama M, Yokoyama S, et al. Total cholesterol performance of Abell-Levy-Brodie-Kendall reference measurement procedure: Certification of Japanese in-vitro diagnostic assay manufacturers through CDC’s Cholesterol Reference Method Laboratory Network. Clin Chim Acta 2015; 445: 127–132.
- 18.
Nakamura M, Iso H, Kitamura A, Imano H, Noda H, Kiyama M, et al. Comparison between the triglycerides standardization of routine methods used in Japan and the chromotropic acid reference measurement procedure used by the CDC Lipid Standardization Programme. Ann Clin Biochem 2016; 53: 632–639.
- 19.
Kirkendall WM, Feinlieb M, Freis ED, Mark AL. Recommendations for human blood pressure determination by sphygmomanometers: Subcommittee of the AHA Postgraduate Education Committee. Circulation 1980; 62: 1146A–1155A.
- 20.
Yaghi S, Bernstein RA, Passman R, Okin PM, Furie KL. Cryptogenic stroke: Research and practice. Circ Res 2017; 120: 527–540.
- 21.
Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation. N Engl J Med 1999; 340: 448–454.
- 22.
Stohs SJ, Bagchi D. Oxidative mechanisms in the toxicity of metal ions. Free Radic Biol Med 1995; 18: 321–336.
- 23.
Jomova K, Valko M. Advances in metal-induced oxidative stress and human disease. Toxicology 2011; 283: 65–87.
- 24.
Neuzil J, Stocker R. Free and albumin-bound bilirubin are efficient co-antioxidants for alpha-tocopherol, inhibiting plasma and low density lipoprotein lipid peroxidation. J Biol Chem 1994; 269: 16712–16719.
- 25.
Purdon AD, Rao AK. Interaction of albumin, arachidonic acid and prostanoids in platelets. Prostaglandins Leukot Essent Fatty Acids 1989; 35: 213–218.
- 26.
Mikhailidis DP, Ganotakis ES. Plasma albumin and platelet function: Relevance to atherogenesis and thrombosis. Platelets 1996; 7: 125–137.
- 27.
Watanabe M, Higashiyama A, Kokubo Y, Ono Y, Okayama A, Okamura T, et al. Protein intakes and serum albumin levels in a Japanese general population: NIPPON DATA90. J Epidemiol 2010; 20: S531–S536.
- 28.
Iso H, Stampfer MJ, Manson JE, Rexrode K, Hu F, Hennekens CH, et al. Prospective study of fat and protein intake and risk of intraparenchymal hemorrhage in women. Circulation 2001; 103: 856–863.
- 29.
Ozawa M, Yoshida D, Hata J, Ohara T, Mukai N, Shibata M, et al. Dietary protein intake and stroke risk in a general Japanese population: The Hisayama Study. Stroke 2017; 48: 1478–1486.
- 30.
Hu H, Sparrow D, Weiss S. Association of serum albumin with blood pressure in the normative aging study. Am J Epidemiol 1992; 136: 1465–1473.
