Fumi Irie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Fumi Irie has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
- Ryu Matsuo
Kyushu University
7 shared publications - Masahiro Kamouchi
Kyushu University
5 shared publications - Tetsuro Ago
Department of Medicine and Clinical Science, Graduate School of Medical Sciences Kyushu university Fukuoka Japan
5 shared publications - Yoshinobu Wakisaka
Department of Medicine and Clinical Science, Graduate School of Medical Sciences Kyushu university Fukuoka Japan
5 shared publications - Takanari Kitazono
Center for Cohort Studies, Graduate School of Medical Sciences, Kyushu University , 3-1-1 Maidashi, Higashi-Ku, Fukuoka 812-8582
4 shared publications Masahiro Kamouchi
Department of Health Care Administration and Management, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan
3 shared publications
Latest Publications
- 1 October 2026
- Published by Wolters Kluwer Health in Stroke
- Vol. 57 (10) , 3046-3054
Independent Association of RNF213 p.R4810K With Extracranial Carotid Artery Disease Beyond Intracranial Atherosclerosis
BACKGROUND: The RNF213 p.R4810K variant, a susceptibility variant for moyamoya disease, has also been linked to intracranial atherosclerotic disease (ICAD). However, its association with extracranial carotid artery disease (ECAD), particularly whether this association is independent of concomitant ICAD and whether it differs by stenosis severity, remains unclear. METHODS: We conducted a multicenter registry-based observational study of patients with acute ischemic stroke (AIS) or transient ischemic attack enrolled in the Fukuoka Stroke Registry. Patients were genotyped for RNF213 p.R4810K and classified as noncarriers (G/G) or variant carriers (G/A or A/A). ECAD was defined as ≥50% stenosis or occlusion of the extracranial carotid arteries. Multivariable logistic regression was used to assess the association between RNF213 p.R4810K and ECAD after adjusting for vascular risk factors and ICAD, and stratified analyses were performed by ICAD status. ECAD severity was modeled ordinally (none, moderate, severe, and occlusion) using a partial proportional-odds model. RESULTS: Of 15 569 registry patients with AIS or transient ischemic attack, 13 994 were included in the present analysis after exclusions (mean age, 73±12 years; 59.5% men). The RNF213 p.R4810K variant was identified in 359 patients (2.6%). ECAD was present in 2181 patients (15.6%), including 78 variant carriers. Variant carriers had significantly higher odds of ECAD after adjustment for vascular risk factors and ICAD (odds ratio, 1.69 [95% CI, 1.29–2.22]). This association was consistent irrespective of ICAD status ( P heterogeneity =0.32). In severity analyses, the association strengthened stepwise with increasing ECAD severity; the largest effect estimate was for occlusion (odds ratio, 4.55 [95% CI, 2.81–7.38]). CONCLUSIONS: The RNF213 p.R4810K variant was associated with ECAD independent of concomitant ICAD and showed a graded association with ECAD severity. Extracranial disease expression may involve modifiers distinct from those driving ICAD.
- 1 October 2026
- Published by Wolters Kluwer Health in Stroke
- Vol. 57 (10) , 3046-3054
Independent Association of RNF213 p.R4810K With Extracranial Carotid Artery Disease Beyond Intracranial Atherosclerosis
BACKGROUND: The RNF213 p.R4810K variant, a susceptibility variant for moyamoya disease, has also been linked to intracranial atherosclerotic disease (ICAD). However, its association with extracranial carotid artery disease (ECAD), particularly whether this association is independent of concomitant ICAD and whether it differs by stenosis severity, remains unclear. METHODS: We conducted a multicenter registry-based observational study of patients with acute ischemic stroke (AIS) or transient ischemic attack enrolled in the Fukuoka Stroke Registry. Patients were genotyped for RNF213 p.R4810K and classified as noncarriers (G/G) or variant carriers (G/A or A/A). ECAD was defined as ≥50% stenosis or occlusion of the extracranial carotid arteries. Multivariable logistic regression was used to assess the association between RNF213 p.R4810K and ECAD after adjusting for vascular risk factors and ICAD, and stratified analyses were performed by ICAD status. ECAD severity was modeled ordinally (none, moderate, severe, and occlusion) using a partial proportional-odds model. RESULTS: Of 15 569 registry patients with AIS or transient ischemic attack, 13 994 were included in the present analysis after exclusions (mean age, 73±12 years; 59.5% men). The RNF213 p.R4810K variant was identified in 359 patients (2.6%). ECAD was present in 2181 patients (15.6%), including 78 variant carriers. Variant carriers had significantly higher odds of ECAD after adjustment for vascular risk factors and ICAD (odds ratio, 1.69 [95% CI, 1.29–2.22]). This association was consistent irrespective of ICAD status ( P heterogeneity =0.32). In severity analyses, the association strengthened stepwise with increasing ECAD severity; the largest effect estimate was for occlusion (odds ratio, 4.55 [95% CI, 2.81–7.38]). CONCLUSIONS: The RNF213 p.R4810K variant was associated with ECAD independent of concomitant ICAD and showed a graded association with ECAD severity. Extracranial disease expression may involve modifiers distinct from those driving ICAD.
- Open Access
- 7 April 2026
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 15 (7) , e045867
Association Between Higher Insulin Resistance and Poor Short‐Term Functional Outcome After Nonsurgical Intracerebral Hemorrhage
Background: Insulin resistance has been implicated in unfavorable outcomes after ischemic stroke in recent studies, but its impact on functional outcomes after intracerebral hemorrhage (ICH) remains unclear. Methods: We analyzed 704 patients with nonsurgical ICH (mean age, 68 years) enrolled in the multicenter Fukuoka Stroke Registry who were treated conservatively and followed for 3 months. Insulin resistance was assessed using the homeostasis model assessment of insulin resistance (HOMA‐IR). There was a poor functional outcome, defined as a modified Rankin Scale score of 4 to 6 at 3 months. Logistic regression and restricted cubic spline analyses were performed. Sensitivity analyses estimated risk ratios and examined results in nondiabetic and nonobese patients. We also assessed the added predictive value of HOMA‐IR when incorporated into the max‐ICH score, a validated prognostic model, and the Fukuoka Stroke Registry ICH score, a developed model from the same registry, using the C statistic and 10‐fold cross‐validation. Results: Age‐ and sex‐adjusted odds ratios (ORs) for poor outcome increased significantly with higher HOMA‐IR ( P for trend 3.0. Adding HOMA‐IR modestly improved discrimination for both risk scores, with similar results in 10‐fold cross‐validation. Conclusions: Insulin resistance is independently associated with poor short‐term outcome after acute ICH. Incorporating HOMA‐IR improves prediction, identifying it as an early prognostic marker and a possible future therapeutic target.
- 7 April 2026
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 15 (7) , e045867
Open AccessAssociation Between Higher Insulin Resistance and Poor Short‐Term Functional Outcome After Nonsurgical Intracerebral Hemorrhage
Background: Insulin resistance has been implicated in unfavorable outcomes after ischemic stroke in recent studies, but its impact on functional outcomes after intracerebral hemorrhage (ICH) remains unclear. Methods: We analyzed 704 patients with nonsurgical ICH (mean age, 68 years) enrolled in the multicenter Fukuoka Stroke Registry who were treated conservatively and followed for 3 months. Insulin resistance was assessed using the homeostasis model assessment of insulin resistance (HOMA‐IR). There was a poor functional outcome, defined as a modified Rankin Scale score of 4 to 6 at 3 months. Logistic regression and restricted cubic spline analyses were performed. Sensitivity analyses estimated risk ratios and examined results in nondiabetic and nonobese patients. We also assessed the added predictive value of HOMA‐IR when incorporated into the max‐ICH score, a validated prognostic model, and the Fukuoka Stroke Registry ICH score, a developed model from the same registry, using the C statistic and 10‐fold cross‐validation. Results: Age‐ and sex‐adjusted odds ratios (ORs) for poor outcome increased significantly with higher HOMA‐IR ( P for trend 3.0. Adding HOMA‐IR modestly improved discrimination for both risk scores, with similar results in 10‐fold cross‐validation. Conclusions: Insulin resistance is independently associated with poor short‐term outcome after acute ICH. Incorporating HOMA‐IR improves prediction, identifying it as an early prognostic marker and a possible future therapeutic target.
- Open Access
- 30 June 2025
- Published by National Center for Global Health and Medicine (JST) in Global Health & Medicine
- Vol. 7 (3) , 233-240
Association between procedure volume and 30-day mortality in stroke patients treated with EVT or IV rt-PA during the introduction period of EVT in Japan
This study aimed to determine whether procedure volume is associated with 30-day mortality following endovascular thrombectomy (EVT) or intravenous recombinant tissue plasminogen activator (IV rt-PA) for stroke during the introduction period of EVT in Japan. Using nationwide claims records, we investigated data from 8,227 patients undergoing EVT and 13,406 and 6,035 patients undergoing rt-PA monotherapy in hospitals with and without EVT capability, respectively, between April 2014 and February 2016 in Japan. Procedure volume was categorized into three groups according to tertiles of the annual number of EVTs or IV rt-PA injections performed in the hospitals. Hierarchical logistic regression demonstrated that the odds ratio (95% confidence interval) of 30-day mortality following EVT was significantly lower in middle- (0.77 [0.62–0.96]) and high- (0.69 [0.53–0.89]) volume hospitals than that in low-volume hospitals even after adjusting for potential confounding factors. The generalized additive mixed models revealed no obvious threshold volume of EVT to reduce the mortality risk. By contrast, mortality risk following IV rt-PA monotherapy did not decrease in hospitals without EVT capability but did with increasing IV rt-PA volume in hospitals with EVT capability (P for heterogeneity 0.003). The risk of 30-day mortality after EVT for acute ischemic stroke decreased linearly according to EVT procedure volume in each hospital. However, the association between IV rt-PA volume and mortality risk was modified by the hospital's EVT capability. Further research is warranted to determine whether the volume-outcome relationship we observed is a temporary phenomenon following EVT or a consistent trend over time.
- 30 June 2025
- Published by National Center for Global Health and Medicine (JST) in Global Health & Medicine
- Vol. 7 (3) , 233-240
Open AccessAssociation between procedure volume and 30-day mortality in stroke patients treated with EVT or IV rt-PA during the introduction period of EVT in Japan
This study aimed to determine whether procedure volume is associated with 30-day mortality following endovascular thrombectomy (EVT) or intravenous recombinant tissue plasminogen activator (IV rt-PA) for stroke during the introduction period of EVT in Japan. Using nationwide claims records, we investigated data from 8,227 patients undergoing EVT and 13,406 and 6,035 patients undergoing rt-PA monotherapy in hospitals with and without EVT capability, respectively, between April 2014 and February 2016 in Japan. Procedure volume was categorized into three groups according to tertiles of the annual number of EVTs or IV rt-PA injections performed in the hospitals. Hierarchical logistic regression demonstrated that the odds ratio (95% confidence interval) of 30-day mortality following EVT was significantly lower in middle- (0.77 [0.62–0.96]) and high- (0.69 [0.53–0.89]) volume hospitals than that in low-volume hospitals even after adjusting for potential confounding factors. The generalized additive mixed models revealed no obvious threshold volume of EVT to reduce the mortality risk. By contrast, mortality risk following IV rt-PA monotherapy did not decrease in hospitals without EVT capability but did with increasing IV rt-PA volume in hospitals with EVT capability (P for heterogeneity 0.003). The risk of 30-day mortality after EVT for acute ischemic stroke decreased linearly according to EVT procedure volume in each hospital. However, the association between IV rt-PA volume and mortality risk was modified by the hospital's EVT capability. Further research is warranted to determine whether the volume-outcome relationship we observed is a temporary phenomenon following EVT or a consistent trend over time.
- Open Access
- 11 January 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (1) , e0296639
Body temperature in the acute phase and clinical outcomes after acute ischemic stroke
Background: This study aimed to examine whether post-stroke early body temperature is associated with neurological damage in the acute phase and functional outcomes at three months. Methods: We included 7,177 patients with acute ischemic stroke within 24 h of onset. Axillary temperature was measured daily in the morning for seven days. Mean body temperature was grouped into five quintiles (Q1: 35.1‒36.5°C, Q2: 36.5‒36.7°C, Q3: 36.7‒36.8°C, Q4: 36.8‒37.1°C, and Q5: 37.1‒39.1°C). Clinical outcomes included neurological improvement during hospitalization and poor functional outcome (modified Rankin scale score, 3–6) at three months. A logistic regression analysis was performed to evaluate the association between body temperature and clinical outcomes. Results: The patient’s mean (SD) age was 70.6 (12.3) years, and 35.7% of patients were women. Mean body temperature was significantly associated with less neurological improvement from Q2 (odds ratios [95% confidence interval], 0.77 [0.65–0.99] vs. Q1) to Q5 (0.33 [0.28–0.40], P for trend <0.001) even after adjusting for potential confounders, including baseline neurological severity, C-reactive protein levels, and post-stroke acute infections. The multivariable-adjusted risk of poor functional outcome linearly increased from Q2 (1.36 [1.03–1.79]) to Q5 (6.44 [5.19–8.96], P for trend 37.0°C. Conclusions: Post-stroke early high body temperature is independently associated with unfavorable outcomes following acute ischemic stroke.
- 11 January 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (1) , e0296639
Open AccessBody temperature in the acute phase and clinical outcomes after acute ischemic stroke
Background: This study aimed to examine whether post-stroke early body temperature is associated with neurological damage in the acute phase and functional outcomes at three months. Methods: We included 7,177 patients with acute ischemic stroke within 24 h of onset. Axillary temperature was measured daily in the morning for seven days. Mean body temperature was grouped into five quintiles (Q1: 35.1‒36.5°C, Q2: 36.5‒36.7°C, Q3: 36.7‒36.8°C, Q4: 36.8‒37.1°C, and Q5: 37.1‒39.1°C). Clinical outcomes included neurological improvement during hospitalization and poor functional outcome (modified Rankin scale score, 3–6) at three months. A logistic regression analysis was performed to evaluate the association between body temperature and clinical outcomes. Results: The patient’s mean (SD) age was 70.6 (12.3) years, and 35.7% of patients were women. Mean body temperature was significantly associated with less neurological improvement from Q2 (odds ratios [95% confidence interval], 0.77 [0.65–0.99] vs. Q1) to Q5 (0.33 [0.28–0.40], P for trend <0.001) even after adjusting for potential confounders, including baseline neurological severity, C-reactive protein levels, and post-stroke acute infections. The multivariable-adjusted risk of poor functional outcome linearly increased from Q2 (1.36 [1.03–1.79]) to Q5 (6.44 [5.19–8.96], P for trend 37.0°C. Conclusions: Post-stroke early high body temperature is independently associated with unfavorable outcomes following acute ischemic stroke.
- Open Access
- 11 January 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (1) , e0296833
Association between abdominal adiposity and clinical outcomes in patients with acute ischemic stroke
Background: It is unclear whether abdominal adiposity has an additional effect on post-stroke outcomes. This study aimed to determine whether waist circumference (WC) is independently associated with clinical outcomes after acute ischemic stroke. Methods: We enrolled patients with acute ischemic stroke from a multicenter hospital-based stroke registry in Fukuoka, Japan. We measured WC on admission and categorized patients into four groups (Q1–Q4) according to the quartiles in females and males. The clinical outcomes were poor functional outcome (modified Rankin scale score 2–6) and death from any cause. Logistic regression analysis was performed to estimate the odds ratio and 95% confidence interval of the outcomes of interest after adjusting for potential confounding factors, including body mass index (BMI). Results: A total of 11,989 patients (70.3±12.2 years, females: 36.1%) were included in the analysis. The risk of poor functional outcome significantly decreased for Q2–Q4 (vs. Q1) at discharge and Q2–Q3 (vs. Q1) at 3 months, even after adjusting for potential confounders, including BMI. In contrast, adjustment of BMI eliminated the significant association between WC and all-cause death at discharge and 3 months. The association between high WC and favorable functional outcome was not affected by fasting insulin levels or homeostatic model assessment for insulin resistance and was only found in patients without diabetes (P = 0.02 for heterogeneity). Conclusions: These findings suggest that abdominal adiposity has an additional impact on post-stroke functional outcome, independent of body weight and insulin action.
- 11 January 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (1) , e0296833
Open AccessAssociation between abdominal adiposity and clinical outcomes in patients with acute ischemic stroke
Background: It is unclear whether abdominal adiposity has an additional effect on post-stroke outcomes. This study aimed to determine whether waist circumference (WC) is independently associated with clinical outcomes after acute ischemic stroke. Methods: We enrolled patients with acute ischemic stroke from a multicenter hospital-based stroke registry in Fukuoka, Japan. We measured WC on admission and categorized patients into four groups (Q1–Q4) according to the quartiles in females and males. The clinical outcomes were poor functional outcome (modified Rankin scale score 2–6) and death from any cause. Logistic regression analysis was performed to estimate the odds ratio and 95% confidence interval of the outcomes of interest after adjusting for potential confounding factors, including body mass index (BMI). Results: A total of 11,989 patients (70.3±12.2 years, females: 36.1%) were included in the analysis. The risk of poor functional outcome significantly decreased for Q2–Q4 (vs. Q1) at discharge and Q2–Q3 (vs. Q1) at 3 months, even after adjusting for potential confounders, including BMI. In contrast, adjustment of BMI eliminated the significant association between WC and all-cause death at discharge and 3 months. The association between high WC and favorable functional outcome was not affected by fasting insulin levels or homeostatic model assessment for insulin resistance and was only found in patients without diabetes (P = 0.02 for heterogeneity). Conclusions: These findings suggest that abdominal adiposity has an additional impact on post-stroke functional outcome, independent of body weight and insulin action.
- Open Access
- 11 January 2024
- Published by JMIR Publications Inc. in JMIR AI
- Vol. 3, e46840
Predictive Performance of Machine Learning–Based Models for Poststroke Clinical Outcomes in Comparison With Conventional Prognostic Scores: Multicenter, Hospital-Based Observational Study
Background: Although machine learning is a promising tool for making prognoses, the performance of machine learning in predicting outcomes after stroke remains to be examined. Objective: This study aims to examine how much data-driven models with machine learning improve predictive performance for poststroke outcomes compared with conventional stroke prognostic scores and to elucidate how explanatory variables in machine learning–based models differ from the items of the stroke prognostic scores. Methods: We used data from 10,513 patients who were registered in a multicenter prospective stroke registry in Japan between 2007 and 2017. The outcomes were poor functional outcome (modified Rankin Scale score >2) and death at 3 months after stroke. Machine learning–based models were developed using all variables with regularization methods, random forests, or boosted trees. We selected 3 stroke prognostic scores, namely, ASTRAL (Acute Stroke Registry and Analysis of Lausanne), PLAN (preadmission comorbidities, level of consciousness, age, neurologic deficit), and iScore (Ischemic Stroke Predictive Risk Score) for comparison. Item-based regression models were developed using the items of these 3 scores. The model performance was assessed in terms of discrimination and calibration. To compare the predictive performance of the data-driven model with that of the item-based model, we performed internal validation after random splits of identical populations into 80% of patients as a training set and 20% of patients as a test set; the models were developed in the training set and were validated in the test set. We evaluated the contribution of each variable to the models and compared the predictors used in the machine learning–based models with the items of the stroke prognostic scores. Results: The mean age of the study patients was 73.0 (SD 12.5) years, and 59.1% (6209/10,513) of them were men. The area under the receiver operating characteristic curves and the area under the precision-recall curves for predicting poststroke outcomes were higher for machine learning–based models than for item-based models in identical populations after random splits. Machine learning–based models also performed better than item-based models in terms of the Brier score. Machine learning–based models used different explanatory variables, such as laboratory data, from the items of the conventional stroke prognostic scores. Including these data in the machine learning–based models as explanatory variables improved performance in predicting outcomes after stroke, especially poststroke death. Conclusions: Machine learning–based models performed better in predicting poststroke outcomes than regression models using the items of conventional stroke prognostic scores, although they required additional variables, such as laboratory data, to attain improved performance. Further studies are warranted to validate the usefulness of machine learning in clinical settings.
- 11 January 2024
- Published by JMIR Publications Inc. in JMIR AI
- Vol. 3, e46840
Open AccessPredictive Performance of Machine Learning–Based Models for Poststroke Clinical Outcomes in Comparison With Conventional Prognostic Scores: Multicenter, Hospital-Based Observational Study
Background: Although machine learning is a promising tool for making prognoses, the performance of machine learning in predicting outcomes after stroke remains to be examined. Objective: This study aims to examine how much data-driven models with machine learning improve predictive performance for poststroke outcomes compared with conventional stroke prognostic scores and to elucidate how explanatory variables in machine learning–based models differ from the items of the stroke prognostic scores. Methods: We used data from 10,513 patients who were registered in a multicenter prospective stroke registry in Japan between 2007 and 2017. The outcomes were poor functional outcome (modified Rankin Scale score >2) and death at 3 months after stroke. Machine learning–based models were developed using all variables with regularization methods, random forests, or boosted trees. We selected 3 stroke prognostic scores, namely, ASTRAL (Acute Stroke Registry and Analysis of Lausanne), PLAN (preadmission comorbidities, level of consciousness, age, neurologic deficit), and iScore (Ischemic Stroke Predictive Risk Score) for comparison. Item-based regression models were developed using the items of these 3 scores. The model performance was assessed in terms of discrimination and calibration. To compare the predictive performance of the data-driven model with that of the item-based model, we performed internal validation after random splits of identical populations into 80% of patients as a training set and 20% of patients as a test set; the models were developed in the training set and were validated in the test set. We evaluated the contribution of each variable to the models and compared the predictors used in the machine learning–based models with the items of the stroke prognostic scores. Results: The mean age of the study patients was 73.0 (SD 12.5) years, and 59.1% (6209/10,513) of them were men. The area under the receiver operating characteristic curves and the area under the precision-recall curves for predicting poststroke outcomes were higher for machine learning–based models than for item-based models in identical populations after random splits. Machine learning–based models also performed better than item-based models in terms of the Brier score. Machine learning–based models used different explanatory variables, such as laboratory data, from the items of the conventional stroke prognostic scores. Including these data in the machine learning–based models as explanatory variables improved performance in predicting outcomes after stroke, especially poststroke death. Conclusions: Machine learning–based models performed better in predicting poststroke outcomes than regression models using the items of conventional stroke prognostic scores, although they required additional variables, such as laboratory data, to attain improved performance. Further studies are warranted to validate the usefulness of machine learning in clinical settings.
- 1 May 2023
- Published by Wolters Kluwer Health in Stroke
- Vol. 54 (5) , 1268-1277
Decreased Estimated Glomerular Filtration Rate and Proteinuria and Long-Term Outcomes After Ischemic Stroke: A Longitudinal Observational Cohort Study
Background: It remains unclear how chronic kidney disease and its underlying pathological conditions, kidney dysfunction, and kidney damage, are associated with cardiovascular outcomes. This study aimed to determine whether kidney dysfunction (ie, decreased estimated glomerular filtration rate), kidney damage (ie, proteinuria), or both are associated with the long-term outcomes after ischemic stroke. Methods: A total of 12 576 patients (mean age, 73.0±12.6 years; 41.3% women) with ischemic stroke who were registered in a hospital-based multicenter registry, Fukuoka Stroke Registry, between June 2007 and September 2019, were prospectively followed up after stroke onset. Kidney function was assessed by estimated glomerular filtration rate and categorized into G1: ≥60 mL/(min·1.73 m 2 ), G2: 45–59 mL/(min·1.73 m 2 ), and G3: 2 ). Kidney damage was evaluated by proteinuria using a urine dipstick test and classified into P1: −, P2: ±/1+, and P3: ≥2+. Hazard ratios and 95% CI for events of interest were estimated by a Cox proportional hazards model. Long-term outcomes included recurrence of stroke and all-cause death. Results: During the median follow-up of 4.3 years (interquartile range, 2.1–7.3 years), 2481 patients had recurrent stroke (48.0/1000 patient-years) and 4032 patients died (67.3/1000 patient-years). Chronic kidney disease was independently associated with increased risks of stroke recurrence and all-cause death even after adjustment for multiple confounding factors, including traditional cardiovascular risk factors. Both estimated glomerular filtration rate and proteinuria were independently associated with increased risks of stroke recurrence (multivariable-adjusted hazard ratio [95% CI], G3: 1.22 [1.09–1.37] versus G1, P3: 1.25 [1.07–1.46] versus P1) and death (G3: 1.45 [1.33–1.57] versus G1, P3: 1.62 [1.45–1.81] versus P1). In subgroup analyses, effect modifications were found in the association of proteinuria with death by age and stroke subtype. Conclusions: Kidney dysfunction and kidney damage were independently, but differently, associated with increased risks of recurrent stroke and all-cause death.
- 1 May 2023
- Published by Wolters Kluwer Health in Stroke
- Vol. 54 (5) , 1268-1277
Decreased Estimated Glomerular Filtration Rate and Proteinuria and Long-Term Outcomes After Ischemic Stroke: A Longitudinal Observational Cohort Study
Background: It remains unclear how chronic kidney disease and its underlying pathological conditions, kidney dysfunction, and kidney damage, are associated with cardiovascular outcomes. This study aimed to determine whether kidney dysfunction (ie, decreased estimated glomerular filtration rate), kidney damage (ie, proteinuria), or both are associated with the long-term outcomes after ischemic stroke. Methods: A total of 12 576 patients (mean age, 73.0±12.6 years; 41.3% women) with ischemic stroke who were registered in a hospital-based multicenter registry, Fukuoka Stroke Registry, between June 2007 and September 2019, were prospectively followed up after stroke onset. Kidney function was assessed by estimated glomerular filtration rate and categorized into G1: ≥60 mL/(min·1.73 m 2 ), G2: 45–59 mL/(min·1.73 m 2 ), and G3: 2 ). Kidney damage was evaluated by proteinuria using a urine dipstick test and classified into P1: −, P2: ±/1+, and P3: ≥2+. Hazard ratios and 95% CI for events of interest were estimated by a Cox proportional hazards model. Long-term outcomes included recurrence of stroke and all-cause death. Results: During the median follow-up of 4.3 years (interquartile range, 2.1–7.3 years), 2481 patients had recurrent stroke (48.0/1000 patient-years) and 4032 patients died (67.3/1000 patient-years). Chronic kidney disease was independently associated with increased risks of stroke recurrence and all-cause death even after adjustment for multiple confounding factors, including traditional cardiovascular risk factors. Both estimated glomerular filtration rate and proteinuria were independently associated with increased risks of stroke recurrence (multivariable-adjusted hazard ratio [95% CI], G3: 1.22 [1.09–1.37] versus G1, P3: 1.25 [1.07–1.46] versus P1) and death (G3: 1.45 [1.33–1.57] versus G1, P3: 1.62 [1.45–1.81] versus P1). In subgroup analyses, effect modifications were found in the association of proteinuria with death by age and stroke subtype. Conclusions: Kidney dysfunction and kidney damage were independently, but differently, associated with increased risks of recurrent stroke and all-cause death.
- Open Access
- 27 February 2023
- Published by JMIR Publications Inc. in JMIR Preprints
Predictive Performance of Machine Learning–Based Models for Poststroke Clinical Outcomes in Comparison With Conventional Prognostic Scores: Multicenter, Hospital-Based Observational Study (Preprint)
BACKGROUND Although machine learning is a promising tool for making prognoses, the performance of machine learning in predicting outcomes after stroke remains to be examined. OBJECTIVE This study aims to examine how much data-driven models with machine learning improve predictive performance for poststroke outcomes compared with conventional stroke prognostic scores and to elucidate how explanatory variables in machine learning–based models differ from the items of the stroke prognostic scores. METHODS We used data from 10,513 patients who were registered in a multicenter prospective stroke registry in Japan between 2007 and 2017. The outcomes were poor functional outcome (modified Rankin Scale score >2) and death at 3 months after stroke. Machine learning–based models were developed using all variables with regularization methods, random forests, or boosted trees. We selected 3 stroke prognostic scores, namely, ASTRAL (Acute Stroke Registry and Analysis of Lausanne), PLAN (preadmission comorbidities, level of consciousness, age, neurologic deficit), and iScore (Ischemic Stroke Predictive Risk Score) for comparison. Item-based regression models were developed using the items of these 3 scores. The model performance was assessed in terms of discrimination and calibration. To compare the predictive performance of the data-driven model with that of the item-based model, we performed internal validation after random splits of identical populations into 80% of patients as a training set and 20% of patients as a test set; the models were developed in the training set and were validated in the test set. We evaluated the contribution of each variable to the models and compared the predictors used in the machine learning–based models with the items of the stroke prognostic scores. RESULTS The mean age of the study patients was 73.0 (SD 12.5) years, and 59.1% (6209/10,513) of them were men. The area under the receiver operating characteristic curves and the area under the precision-recall curves for predicting poststroke outcomes were higher for machine learning–based models than for item-based models in identical populations after random splits. Machine learning–based models also performed better than item-based models in terms of the Brier score. Machine learning–based models used different explanatory variables, such as laboratory data, from the items of the conventional stroke prognostic scores. Including these data in the machine learning–based models as explanatory variables improved performance in predicting outcomes after stroke, especially poststroke death. CONCLUSIONS Machine learning–based models performed better in predicting poststroke outcomes than regression models using the items of conventional stroke prognostic scores, although they required additional variables, such as laboratory data, to attain improved performance. Further studies are warranted to validate the usefulness of machine learning in clinical settings.
- 27 February 2023
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessPredictive Performance of Machine Learning–Based Models for Poststroke Clinical Outcomes in Comparison With Conventional Prognostic Scores: Multicenter, Hospital-Based Observational Study (Preprint)
BACKGROUND Although machine learning is a promising tool for making prognoses, the performance of machine learning in predicting outcomes after stroke remains to be examined. OBJECTIVE This study aims to examine how much data-driven models with machine learning improve predictive performance for poststroke outcomes compared with conventional stroke prognostic scores and to elucidate how explanatory variables in machine learning–based models differ from the items of the stroke prognostic scores. METHODS We used data from 10,513 patients who were registered in a multicenter prospective stroke registry in Japan between 2007 and 2017. The outcomes were poor functional outcome (modified Rankin Scale score >2) and death at 3 months after stroke. Machine learning–based models were developed using all variables with regularization methods, random forests, or boosted trees. We selected 3 stroke prognostic scores, namely, ASTRAL (Acute Stroke Registry and Analysis of Lausanne), PLAN (preadmission comorbidities, level of consciousness, age, neurologic deficit), and iScore (Ischemic Stroke Predictive Risk Score) for comparison. Item-based regression models were developed using the items of these 3 scores. The model performance was assessed in terms of discrimination and calibration. To compare the predictive performance of the data-driven model with that of the item-based model, we performed internal validation after random splits of identical populations into 80% of patients as a training set and 20% of patients as a test set; the models were developed in the training set and were validated in the test set. We evaluated the contribution of each variable to the models and compared the predictors used in the machine learning–based models with the items of the stroke prognostic scores. RESULTS The mean age of the study patients was 73.0 (SD 12.5) years, and 59.1% (6209/10,513) of them were men. The area under the receiver operating characteristic curves and the area under the precision-recall curves for predicting poststroke outcomes were higher for machine learning–based models than for item-based models in identical populations after random splits. Machine learning–based models also performed better than item-based models in terms of the Brier score. Machine learning–based models used different explanatory variables, such as laboratory data, from the items of the conventional stroke prognostic scores. Including these data in the machine learning–based models as explanatory variables improved performance in predicting outcomes after stroke, especially poststroke death. CONCLUSIONS Machine learning–based models performed better in predicting poststroke outcomes than regression models using the items of conventional stroke prognostic scores, although they required additional variables, such as laboratory data, to attain improved performance. Further studies are warranted to validate the usefulness of machine learning in clinical settings.
- Open Access
- 23 September 2022
- Published by Food Safety Commission, Cabinet Office in Food Safety
- Vol. 10 (3) , 102-111
Commentary on the Risk Assessment of Lead by the Food Safety Commission of Japan
This article describes in detail the process of and the basis for the risk assessment of lead, started as a self-tasking assessment in April 2008 and finalized in June 2021 by the Food Safety Commission of Japan (FSCJ). Discussion points addressed in the working group set under the FSCJ in April 2019 are also presented in this commentary. To reflect the overall exposure to lead from various sources, blood lead level (BLL) was used as the basic metric for the assessment. For the evaluation of effects on human health, the approach of overall weight of evidence was taken, rather than selecting one critical endpoint, in consideration of the uncertainties inherent to epidemiological studies, particularly those examining the effects associated with low-level lead exposure. The overall evidence compiled for the assessment suggested that BLLs in the range of 1–2 μg/dL might be associated with some effects on human health. The representative value of BLL for the entire population was difficult to obtain due to the lack of a national population-based survey in Japan. Instead, the current average BLL of the Japanese population was estimated based on recent studies conducted in Japan. The estimated average exposure level was below or equal to 1 μg/dL and close to the levels at which some effects on human health might occur, as suggested by epidemiological studies. Hence, the continued enforcement of measures to reduce lead exposure is indispensable. Furthermore, a national human biomonitoring program to continuously assess the exposure status of the Japanese population, which can be ultimately used for assuring the effectiveness of control measures, is needed.
- 23 September 2022
- Published by Food Safety Commission, Cabinet Office in Food Safety
- Vol. 10 (3) , 102-111
Open AccessCommentary on the Risk Assessment of Lead by the Food Safety Commission of Japan
This article describes in detail the process of and the basis for the risk assessment of lead, started as a self-tasking assessment in April 2008 and finalized in June 2021 by the Food Safety Commission of Japan (FSCJ). Discussion points addressed in the working group set under the FSCJ in April 2019 are also presented in this commentary. To reflect the overall exposure to lead from various sources, blood lead level (BLL) was used as the basic metric for the assessment. For the evaluation of effects on human health, the approach of overall weight of evidence was taken, rather than selecting one critical endpoint, in consideration of the uncertainties inherent to epidemiological studies, particularly those examining the effects associated with low-level lead exposure. The overall evidence compiled for the assessment suggested that BLLs in the range of 1–2 μg/dL might be associated with some effects on human health. The representative value of BLL for the entire population was difficult to obtain due to the lack of a national population-based survey in Japan. Instead, the current average BLL of the Japanese population was estimated based on recent studies conducted in Japan. The estimated average exposure level was below or equal to 1 μg/dL and close to the levels at which some effects on human health might occur, as suggested by epidemiological studies. Hence, the continued enforcement of measures to reduce lead exposure is indispensable. Furthermore, a national human biomonitoring program to continuously assess the exposure status of the Japanese population, which can be ultimately used for assuring the effectiveness of control measures, is needed.