Table 13 Studies examining the relationship between physical activity and stroke.

Wisloff et al 2006 [58]

To assess exercise amount and intensity in relation to subsequent CVD mortality (including stroke).

• n = 27,143 men, 28,929 women

16 year follow up

Multivariate adjusted RR (95% CI) Men

Both high and low- intensity exercise may be associated with a reduced risk of stroke in both men and women.

Norway

• Sex: Men and women

PA Assessment: Questionnaire

G1 = 1.00 (referent)

• Age: ≥ 20 yr

G2 = 0.90 (0.70-1.17)

• Characteristics: free from CVD

PA

G3a = 0.90 (0.64-1.26)

• HUNT Study

G1 = None

G3b = 0.59 (0.27-1.27)

G2 = <1/wk

G3c = 0.62 (0.40-0.95)

G3a = 1/wk ≤ 30 min low

G3d = 0.51 (0.31-0.86)

G3b = 1/wk ≤ 30 min high

G4a = 0.72 (0.49-1.05)

G3c = 1/wk > 30 min low

G4b = 0.63 (0.31-1.30)

Prospective cohort

G3d = 1/wk > 30 min high

G4c = 1.02 (0.72-1.44)

G4a = 2-3/wk ≤ 30 min low

G4d = 0.59 (0.37-0.92)

G4b = 2-3/wk ≤ 30 min high

G5a = 0.97 (0.70-1.36)

D & B score = 12

G4c = 2-3/wk > 30 min low

G5b = 0.68 (0.27-1.66)

G4d = 2-3/wk > 30 min high

G5c = 0.81 (0.65-1.20)

G5a = ≥ 4/wk ≤ 30 min low

G5d = 0.67 (0.49-1.11)

G5b = ≥ 4/wk ≤ 30 min high

G5c = ≥ 4wk > 30 min low

RR (95% CI) Women

G5d = ≥ 4/wk > 30 min high

G1 = 1.00 (referent)

Outcome Measure: IHD mortality

G2 = 1.01 (0.81-1.25)

Cox proportional HR

G3a = 0.88 (0.68-1.15)

G3b = 0.98 (0.46-2.10)

G3c = 0.63 (0.42-0.94)

G3d = 1.00 (0.50-1.98)

G4a = 0.91 (0.70-1.17)

G4b = 1.44 (0.78-2.65)

G4c = 0.62 (0.44-0.88)

G4d = 0.77 (0.36-1.66)

G5a = 0.74 (0.56-0.99)

G5b = 0.40 (0.10-1.62)

G5c = 0.63 (0.45-0.89)

G5d = 0.51 (0.21-1.26)

Abbott et al 2003 [69]

To examine the way in which risk factor effects on the incidence of thromboembolic and hemorrhagic stroke can change over a broad range of ages.

• n = 7,589

6, 15 and 26 year follow up

Incidence rates per 1000 of stroke:

The protective effect of PA on reducing risk of stroke increased with age.

USA

• Sex: Men

   • G1 = 9.0 (49)

• Age: 45-93 yr

PA assessment: Using PA index over a 24 hour period PA information collected at study enrolment 1965-1968 and updated at physical examinations that occurred at 6, 15 and 26 years into follow-up.

   • G2 = 17.8 (124)

Prospective cohort

• Characteristics: Free from CHD and stroke at enrolment; Japanese ancestry living on the island of Oahu, Hawaii.

Grouped into 4 age groups, yr:

   • G3 = 33.4 (112)

D & B score = 14

• Honolulu Heart Program

G1 = 45-54

   • G4 = 48.1 (111)

G2 = 55-64

Incidence of stroke event increased with advancing age p <0.001

G3 = 65-74

There appeared to be a small protective effect within each age group. Inverse relations increased with age (p = 0.046). The protective effect of PA became significant in men >77 years (p = 0.032)

G4 = 75-93

Outcome Measure: diagnosis of fatal and non fatal stroke during 26 years of follow-up

Cox proportional HR

Gillium et al 1996 [70]

To examine the relationship between recreational and non-recreational PA and risk of stroke.

• n = 2,368 men, 2,713 women

11.6 year follow up

Number of Cases: 249 white women, 270 white men, 104 black

Sedentary behaviour was found to be associated with increased risk of stroke.

USA

• Sex: Men and women

PA assessment: Questionnaire divided into tertiles:

• Age: 45-74 yr

T1 = Low

RR (95% CI) Black men and women Recreational PA

Prospective cohort

• Ethnicity: Black and white

T2 = Medium

   • T1 = 1.33 (0.67-2.63)

D & B score = 12

• NHANES I

T3 = High

   • T2 = 1.33 (0.63-2.79)

   • T3 = 1.00 (referent)

Outcome Measure: Total Stroke

Non-recreational PA

Cox proportional HR

   • T1 = 1.40 (0.90-2.16)

   • T2 = 1.41 (0.74-2.70)

   • T3 = 1.00 (referent)

RR (95% CI) White men age 45-64 Recreational PA

   • T1 = 1.24 (0.63-2.41)

   • T2 = 1.17 (0.61-2.27

   • T3 = 1.00 (referent)

Non-recreational PA

   • T1 = 1.07 (0.40-2.86)

   • T2 = 1.75 (1.04-2.96)

   • T3 = 1.00 (referent)

RR (95% CI) White women age 45-64 Recreational PA

   • T1 = 3.13 (0.95-10.32)

   • T2 = 1.80 (0.52-6.22)

   • T3 = 1.00 (referent)

Non-recreational PA

   • T1 = 3.51 (1.66-7.46)

   • T2 = 1.07 (0.57-1.99)

   • T3 = 1.00 (referent)

RR (95% CI) White men age 65-74 Recreational PA

   • T1 = 1.29 (0.58-1.88)

   • T2 = 0.86 (0.58-1.28)

   • T3 = 1.00 (referent)

Non-recreational

   • T1 = 1.82 (1.15-2.88)

   • T2 = 1.20 (0.88-1.64)

   • T3 = 1.00 (referent)

RR (95% CI) White women age 65-75 Recreational PA

   • T1 = 1.55 (0.95-2.53)

   • T2 = 1.27 (0.76-2.12)

   • T3 = 1.00 (referent)

Non-recreational PA

   • T1 = 1.82 (1.10-3.02)

   • T2 = 1.42 (1.01-2.00)

   • T3 = 1.00 (referent)

Lee and Blair 2002 [71]

To examine the association between PF and stroke mortality in men.

• n = 16,878

Baseline medical evaluation between 1971 and 1994 with average follow up period of 10 years

Average estimated maximal METs

Moderate and high levels of PF were associated with lower risk of stroke mortality in men.

• Sex: Men

   • T1 = 8.5 MET

• Age: 40-87 yrs

   • T2 = 10.5 MET

USA

• Aerobics Center Longitudinal Study

   • T3 = 13.1 MET

Prospective cohort

PF assessment: Maximal exercise tolerance test, divided into tertiles

RR (95% CI) adjusted for age and exam year

   • T1 = 1.00 (referent)

D & B score = 13

T1 = Low

   • T2 = 0.35 (0.16-0.77)

T2 = Moderate

   • T3 = 0.28 (0.11-0.71)

T3 = High

Trend p = 0.005

Cox proportional HR

Hu et al 2000 [72]

To examine the association between PA and risk of total stroke and stroke sub- types in women.

• n = 72,488

Baseline measurement in 1986 with follow-up questionnaire in 1988 and 1992

• 407 cases of stroke (258 ischemic strokes, 67 subarachnoid hemorrhages, 42 intracerebral hemorrhages, and 40 strokes of unknown type)

PA, including moderate-intensity exercise such as walking, is associated with a substantial reduction in risk of total and ischemic stroke in a dose- response manner.

• Sex: Women

• Age:40-65 yr

USA

• Characteristics: Nurses

Prospective cohort

• Nurses' Health Study

PA assessment: Questionnaire for total PA (MET h/wk), divided into quintiles, walking activity (MET h/wk), divided into quintiles and walking pace

Multivariate RR (95% CI) for total stroke by total PA level

   • Q1 = 1.00 (referent)

D & B score = 13

   • Q2 = 0.98

   • Q3 = 0.82

   • Q4 = 0.74

   • Q5 = 0.66

Total PA (MET h/wk)

p = 0.005

Q1 = 0 - 2.0

Q2 = 2.1 - 4.6

Multivariate RR (95% CI) for ischemic Stroke by total PA level

Q3 = 4.7 - 10.4

Q4 = 10.5-21.7

   • Q1 = 1.00 (referent)

Q5 = > 21.7

   • Q2 = 0.87

Walking activity (MET h/wk)

   • Q3 = 0.83

Q1 = 0.5

   • Q4 = 0.76

Q2 = 0.6 - 2.0

   • Q5 = 0.52

Q3 = 2.1 - 3.8

p = 0.003

Q4 = 3.9 - 10

Q5 = 10

Multivariate RR (95% CI) for total stroke by walking activity

Walking pace (mph)

   • Q1 = 1.00 (referent)

G1 < 2.0

   • Q2 = 0.76

G2 = 2-2.9

   • Q3 = 0.78

G3 3.0

   • Q4 = 0.70

   • Q5 = 0.66

Outcome measure: Stroke incidence

p = 0.01

Multivariate RR (95% CI) for ischemic stroke by walking activity

Pooled logistic regression

Cox proportional HR

   • Q1 = 1.00 (referent)

   • Q2 = 0.77

   • Q3 = 0.75

   • Q4 = 0.69

   • Q5 = 0.60

p = 0.02

Multivariate RR (95% CI) for total stroke by usual Walking Pace

   • G1 = 1.00 (referent)

   • G2 = 0.81

   • G3 = 0.49

p < 0.001

Multivariate RR (95% CI) for ischemic stroke by usual walking pace

   • G1 = 1.00 (referent)

   • G2 = 0.71

   • G3 = 0.47

p < 0.001

Lee et al 1999 [74]

To examine the association between exercise and stroke risk.

• n = 21,823

11.1 year follow up

Number of Cases: 533

VPA is associated with a decreased risk of stroke in men.

• Sex: Men

• Age: 40-84 yr

PA assessment: Questionnaire for frequency of VPA, divided into 4 groups

Multivariate RR1 (95% CI) for total stroke by VPA

USA

   • G1 = 1.00 (referent)

Prospective cohort

   • G2 = 0.79 (0.61-1.03)

Inverse association with PA seemed to be mediated through beneficial effects on body weight, BP, cholesterol and glucose tolerance.

G1 < 1 time/week

   • G3 = 0.80 (0.65-0.99)

G2 = 1 time/week

   • G4 = 0.79 (0.61-1.03)

D & B score = 13

G3 = 2-4 times/week

p = 0.04

G4 ≥ 5 times/week

RR2 (95% CI) for total stroke by VPA

   • G1 = 1.00 (referent)

RR1 = adjusted for smoking, alcohol consumption, history of angina and parental history of MI at <60 years

   • G2 = 0.81 (0.61-1.07)

   • G3 = 0.88 (0.70-1.10)

   • G4 = 0.86 (0.65-1.13)

p = 0.25

RR2 (95% CI) for ischemic stroke by

RR2 = adjusted for all of the above plus, BMI, history of, hypertension, high cholesterol and diabetes

VPA

   • G1 = 1.00 (referent)

   • G2 = 0.90 (0.66-1.22)

   • G3 = 0.95 (0.74-1.22)

   • G4 = 0.97 (0.71-1.32)

Outcome Measure: Total Stroke (Ischemic and Hemorrhagic)

p = 0.81

RR2 (95% CI) for hemorrhagic stroke by VPA

Cox proportional HR

   • G1 = 1.00 (referent)

   • G2 = 0.54 (0.25-1.13)

   • G3 = 0.71 (0.41-1.23)

   • G4 = 0.54 (0.26-1.15)

p = 0.10

Bijnen et al 1998 [166]

To describe the association between the PA patterns of elderly men and stroke mortality.

• n = 802

10 year follow up

Number of Cases: 47

No significant finding

• Sex: Men

• Age:64-84 yr

PA assessment:

Multivariate adjusted RR (95% CI)

Denmark

• Characteristics: Not all free from previous stroke

Questionnaire for LTPA, divided into tertiles

   • T1= 1. 00 (referent)

   • T2 = 0.65 (0.33-1.25)

Prospective cohort

T1 = Lowest

   • T3 = 0.55 (0.24-1.26)

T2

p = 0.12

T3 = Highest

D & B score = 15

Outcome Measure: Stroke Mortality

Cox proportional HR

Schnohr et al 2006 [214]

To describe the association between different levels of LTPA and subsequent causes of death (stroke).

• n = 2136 men, 2,758 women

5 year follow up

RR (95% CI), univariate

Although RR for of death from stroke was below 1 for both moderate and high compared with low PA, this association did not reach the level of statistical significance.

   • G1 = 1.00 (referent)

• Sex: Men and women

PA assessment:

   • G2 = 0.64 (0.39-1.05)

Copenhagen

• Age: 20 -- 79 yr

Questionnaire for LTPA,

   • G3 = 0.70 (0.41-1.21)

• Characteristics: Healthy, PA level did not change between 2 examinations, 5 years apart

divided into 3 groups

Trend p = 0.4

Prospective cohort

G1 = Low PA (<4 METS)

G2 = Moderate PA (4-6

RR (95% CI), multivariate:

METS)

   • G1 = 1.00 (referent)

D & B score = 13

G3 = High PA (>6 METS)

   • G2 = 0.67 (0.40-1.12)

• Copenhagen City Heart Study

   • G3 = 0.76 (0.43-1.34)

Multivariate Analysis Kaplan-Meier Plots

Trend p = 0.6

Linear, Logistical and Cox Regression.

Vatten et al 2006 [253]

To investigate whether obesity- related CV mortality could be modified by PA.

• n = 26,515 men, 27,769 women

16 year follow up

Number of Cases: 994 women, 771 men

Lower levels of TPA are associated with an increased risk of stroke.

• Sex: Men and women

PA assessment: Questionnaire for total amount of PA, divided into 4 groups

Norway

• Age: 20 yr

Multivariate HR (95% CI), men

• Characteristics: Free from CVD at baseline

   • Q1 = 1.00 (referent)

Prospective cohort

   • Q2 = 1.05 (0.85-1.30)

• HUNT study

G1 = High

   • Q3 = 1.21 (0.95-1.54)

G2 = medium

   • Q4 = 1.35 (1.05-1.74)

D & B score = 14

G3 = low

p = 0.009

G4 = never

Multivariate HR (95% CI), women

Outcome Measure: Stroke mortality

   • Q1 = 1.00 (referent)

   • Q2 = 1.16 (0.93-1.45)

   • Q3 = 1.45 (1.14-1.86)

Cox proportional HR

   • Q4 = 1.45 (1.14-1.83)

p < 0.001

Agnarsson et al 1999 [255]

To examine the association of LTPA and pulmonary function with the risk of stroke.

• n = 4,484

Length of Follow-up: 10.6 ± 3.6 years

Number of Cases: 249

Apparent protective effect of regular continued LTPA in middle age men on the risk of ischemic stroke.

• Sex: Men

• Age: 45-80

Adjusted for age and smoking RR (95% CI) for total stroke by LTPA level

Iceland

• Characteristics: no history of Stroke

PA assessment: Questionnaire for LTPA (h/wk) and type of activity (intensity), each divided into 3 groups

Prospective cohort

• Reykjavik Study

   • G1 = 1.00 (referent)

   • G2 = 0.84 (0.63-1.13)

   • G3 = 0.73 (0.40-1.35)

D & B score = 13

LTPA summer/winter

G1 = none

Adjusted for age and smoking RR (95% CI) for ischemic stroke by LTPA level

G3 = ≥ 6 h/wk

   • G1 = 1.00 (referent)

Type of Activity

   • G2 = 0.72 (0.51-1.01)

G1 = none

   • G3 = 0.78 (0.41-1.48)

G2 = low intensity

G3 = high Intensity

RR (95% CI) for total stroke by type of activity

Outcome Measure: Total and ischemic Stroke

   • G1 = 1.0,0 (referent)

   • G2 = 0.75 (0.53-1.08)

   • G3 = 1.10 (0.78-1.57)

Cox proportional HR

RR (95% CI) for ischemic stroke by type of activity

   • G1 = 1.00 (referent)

   • G2 = 0.72 (0.44-1.07)

   • G3 = 0.96 (0.64-1.44)

Ellekjaer et al 2000 [256]

To examine the association between different levels of LTPA and stroke mortality in middle-aged and elderly women.

• n = 14,101

Baseline 1984-1986: 2 self administered questionnaires and clinical measurements included in the screening program.

Number of cases: 457

This study demonstrates a consistent, negative association between PA and stroke mortality in women.

• Sex: Women

• Age: 50 yr

Multivariate RR (95% CI), all age groups

Norway

• Characteristics: free from stroke at baseline

   • G1 = 1.00 (referent)

Prospective cohort

   • G2 = 0.77

PA assessment: Questionnaire for LTPA, divided into 3 groups

   • G3 = 0.52

D & B score = 14

Multivariate RR (95% CI), age 50--69 years

G1 = low

The most active women had approx. 50% lower risk of death from stroke compare to inactive women.

G2 = medium

   • G1 = 1.00 (referent)

G3 = high

   • G2 = 0.57

   • G3 = 0.42

Outcome Measure: Death from stroke

p = 0.0021

Multivariate RR (95% CI), age 70-79 years

Cox proportional HR

   • G1 = 1.00 (referent)

   • G2 = 0.79

   • G3 = 0.56

p = 0.0093

Multivariate RR (95% CI), age 80-101 years

   • G1 = 1.00 (referent)

   • G2 = 0.91

   • G3 = 0.57

p = 0.1089

Evenson et al 1999 [257]

To examine the relationship between PA and ischemic stroke risk.

• n = 14,575

7.2 year follow up

Number of Cases: 189

PA was weakly associated with a reduced risk of ischemic stroke among middle aged adults.

• Sex: Men and women

Number of Dropouts: 0%

• Age: 45-64 yr

PA assessment: Questionnaire (Baecke questionnaire)

USA

• Atherosclerosis Risk in Communities Study

Sport, Incidence of Ischemic Stroke

Prospective cohort

Multivariate adjusted RR (95% CI) by sport

Outcome Measure:

Ischemic Stroke

   • Q1 = 1.00 (referent)

D & B score = 14

   • Q3= 0.83 (0.52-1.32)

Multivariate Poisson and Cox proportional HR

Multivariate adjusted RR (95% CI) by LTPA

   • Q1 = 1.00 (referent)

   • Q2 =

   • Q3 = 0.89 (0.57-1.37)

Multivariate adjusted RR (95% CI) by OPA

   • Q1 = 1.00 (referent)

   • Q2 =

   • Q3 = 0.69 (0.47-1.00)

Haheim et al 1993 [258]

To determine the risk factors of stroke incidence and mortality.

• n = 14,403

Baseline Screening from May 1972- December 1973.

HR (95% CI) for stroke incidence

Increased LTPA is associated with a reduced risk of stroke incidence but not mortality.

• Sex: Men

   • G1 = 1.00 (referent)

• Age: 40-49 yr

   • G2 = 0.64 (0.38-1.08)

Norway

PA assessment: Questionnaire for LTPA, divided into groups

   • G3 = 0.36 (0.15-0.80)

Prospective cohort

HR (95% CI) for stroke mortality

G1 = Sedentary

   • G1 = 1.00, (referent)

G2 = Moderate

   • G2 = 0.82 (0.33-2.35)

D & B score = 14

G3 = Intermediate or Great

   • G3 = 0.29 (0.03-1.51)

Outcome Measure: Incidence of stroke morbidity and mortality until study end date, December 31, 1984.

Cox proportional HR

Hu et al 2005 [259]

To assess the relationship of different types of PA with total and type-specific stroke risk.

• n = 47,721

PA assessement: Mailed questionnaire for LTPA, OPA and commuting PA, divided into groups as follows:

RR (95% CI) by LTPA, men

A high level of LTPA reduces the risk of all subtypes of stroke. Daily active commuting also reduces the risk of ischemic stroke.

• Sex: Men and women

   • G1 = 1.00 (referent)

• G2 = 0.83

Finland

• Age: 25-64

   • G3 = 0.72

• Characteristics: Healthy at baseline

p < 0.001

Prospective cohort

LTPA levels:

RR (95% CI) by LTPA, women

G1 = Low

   • G1 = 1.00 (referent)

D & B score = 13

G2 = Moderate

   • G2 = 0.86

G3 = High

   • G3 = 0.75

p = 0.007

OPA:

G1 = Light

RR (95% CI) by LTPA, men and women

G2 = Moderate

G3 = Hard

   • G1 = 1.00 (referent)

   • G2 = 0.85

Commuting PA:

G1 = Motorized or no work,

   • G3 = 0.73

G2 = walking or cycling 1-29 min G3 = walking or cycling ≥ 30 min.

p <0.001

RR (95% CI) by OPA, men

• Not significant

Outcome Measure: Incidence of fatal or non-fatal stroke occurring during follow-up until end of 2003. Mean follow-up of 19 years.

RR (95% CI) by OPA, women

• Not significant

RR (95% CI) by OPA, men and women

   • G1 = 1.00 (referent)

Cox proportional hazard

   • G2 = 0.90

• G3 = 0.87

p = 0.007

RR (95% CI) by commuting PA, men

   • G1 = 1.00 (referent)

   • G2 = 0.91

   • G3 = 0.85

p = 0.047

RR (95% CI) by commuting PA, women

   • G1 = 1.00 (referent)

   • G2 = 0.86

   • G3 = 0.85

p = 0.018

RR (95% CI) by commuting PA, men and women

   • G1 = 1.00 (referent)

   • G2 = 0.89

   • G3 = 0.85

p = 0.002

Kiely et al 1994 [260]

To examine the influence of increased PA on stroke risk in members of the Framingham study cohort.

• n = 1,897 men 2,299 women

Baseline measurement in 1954-1955 and follow up in either 1968-1969 or 1971- 1972

Multivariate adjusted RR (95% CI) at first examination, men (mean age 50 years)

Medium and high levels of PA among men are protective against stroke relative to low levels.

• Sex: Men and women

USA

   • G1 = 1.00 (referent)

• Age: 28-62 yr

   • G2 = 0.90 (0.62-1.31) p = 0.59

Prospective cohort

• Characteristics: Free from stroke

PA assessment: Questionnaire for metabolic work done during a typical 24 hr period, divided into 3 groups

   • G3 = 0.84 (0.59-1.18) p = 0.31

Multivariate adjusted RR (95% CI) at first examination, women (mean age 50 years)

Protective effect of PA was slightly less for high levels of PA compared to medium levels for older men.

D & B score = 12

G1 = Low

   • G1 = 1.00 (referent)

G2 = Medium

   • G2 = 1.21 (0.89-1.63) p = 0.23

G3 = High

   • G3 = 0.89 (0.60-1.31) p = 0.54

Outcome Measure: Incidence of stroke, as defined by the first occurrence of atherothrombotic brain infarctions, cerebral embolism or other type of stroke, during 32 years of follow-up.

Multivariate adjusted RR (95% CI) at second examination, men (mean age 63 years)

   • G1 = 1.00 (referent)

   • G2 = 0.41 (0.24-0.89) p = 0.0007

   • G3 = 0.53 (0.34-0.84) p = 0.007

Multivariate adjusted RR (95% CI) at second examination, women (mean age 64 years)

Cox proportional HR

   • G1 = 1.00 (referent)

   • G2 = 0.97 (0.64-1.47) p = 0.67

   • G3 = 1.21 (0.75-1.96) p = 0.43

Krarup et al 2007 [261]

To compare the reported level of PA performed during the week preceding an ischemic stroke with that of community controls.

• n = 127 cases 301 controls

PA assessment:

Univariate OR (95% CI)

Stroke patients are less physically active in the week preceding an ischemic stroke when compared to age and sex-matched controls. Increasing PASE score was inversly, log-linearly and significantly associated with OR for ischemic stroke.

Questionnaire about PA 1 week prior to stroke (cases) and 1 week prior to questionnaire (controls), divided into PASE scores and quartiles

PASE Score

• Sex: Men and women

   • Q1 = 1.00 (referent)

Denmark

   • Q2 = 0.51 (0.28-0.95)

• Age: ≥ 40 yr

• Q3 = 0.27 (0.14-0.54)

Case control

• Characteristics: Case: Stroke Patients (20% had history of Stroke), Controls: 4% had history of stroke

• Q4 = 0.08 (0.03-0.20)

D & B score = 14

Q1 = 0-49

Multivariate OR (95% CI) PASE Score

Q2 = 50-99

Q3 = 100-149

   • Q1 = 1.00 (referent)

Q4 = 150+

   • Q2 = 0.53 (0.26-1.08)

   • Q3 = 0.27 (0.12-0.59)

Outcome measure:

Ischemic stroke

   • Q4 = 0.09 (0.03-0.25)

Chi squared Kruskal-Wallis Statistics Multivariate conditional logistic regression

Kurl et al 2003 [262]

To examine the relationship of PF with subsequent incidence of stroke. Also to compare PF with conventional risk factors as a predictor for future stroke.

• n = 2,011

Baseline examinations conducted between March 1984 and December 1989 with average follow up period of 11 years

Multivariate HR (95% CI), any stroke

Low PF was associated with an increased risk of any stroke and ischemic stroke.

• Sex: Men

   • Q1 = 1.00 (referent)

• Age: 42, 48, 54 or 60 yrs

   • Q2 = 1.39 (0.70-2.77)

Finland

   • Q3 = 1.32 (0.66-2.65)

• Characteristics: Free from stroke or pulmonary disease

• Kuopio Ischaemic Heart Disease Risk Factor Study

   • Q4 = 2.30 (1.18-4.06)

Prospective cohort

Trend p = 0.01

PF assessment: Maximal exercise test on cycle ergometer. VO₂ max (ml/kg/min) divided into quartiles

Multivariate HR (95% CI), ischemic stroke

D & B score = 14

   • Q1 = 1.00 (referent)

   • Q2 = 1.28 (0.56-2.94)

   • Q3 = 1.64 (0.74-3.65)

Q1 = >35.3

   • Q4 = 2.40 (1.09-5.25)

Q2 = 30.3-35.3

Trend p = 0.01

Q3 = 25.2-30.2

Q4 = <25.2

Outcome Measure: Stroke incidence

Cox proportional HR

Myint et al 2006 [263]

To examine the association between a combination of OPA and LTPA with risk of subsequent stroke.

• n = 22,602

Baseline measurement in

Model A: Used all 4 categories of PA

Higher levels of PA assessed using a single simple pragmatic tool based on both OPA and LTPA is associated with reduced stroke risk.

• Sex: Men

1993-1997

HR (95% CI), men and women

• Age: 40-79 yr

   • G1 = 1.00 (referent)

UK

• Characteristics: Healthy at baseline

PA assessment: Questionnaire for PA (includes LTPA and OPA) divided into 4 groups

   • G2 = 0.78 (0.61-1.00)

   • G3 = 0.66 (0.49-0.91)

Prospective cohort

• European Prospective Investigation in Cancer-Norfolk

   • G4 = 0.70 (0.49-0.99)

p = 0.024

D & B score = 11

G1 = Inactive

HR (95% CI), men

G2 = moderately inactive

   • G1 = 1.00 (referent)

G3 = moderately active

   • G2 = 0.75 (0.52-1.09)

G4 = active

   • G3 = 0.55 (0.35-0.86)

   • G4 = 0.67 (0.43-1.05)

Outcome Measure: Incidence of fatal and non fatal stroke.

p = 0.41

Women not significant p = 0.50

Cox proportional HR

Model B: Used 3 categories of PA (G3 and G4 combined combined)

HR (95% CI), men and women

   • G1 = 1.00 (referent)

   • G2 = 0.78 (0.61-1.00)

   • G3 = 0.68 (0.52-0.88)

p = 0.009

HR (95% CI), men

   • G1 = 1.00 (referent)

   • G2 = 0.75 (0.52-1.09),

   • G3 = 0.61 (0.43-0.86)

p = 0.019

Women not significant p = 0.34

Noda et al 2005 [264]

To examine the impact of exercise on CVD (stroke) mortality in Asian populations.

• n = 31,023 men, 42,242 women

9.7 year follow up

Number of Cases: 186 men, 141 women

PA through walking and sports participation may reduce the risk of mortality from ischemic stroke

• Sex: Men and women

PA assessment: Questionnaire for PA (walking and sports participation (h/day), divided into quartiles:

Number of Dropouts: 3.4%

Japan

• Age: 40 -79 yr

• Ethnicity: Asian

Multivariate adjusted HR (95% CI) by duration of walking PA, men

Prospective cohort

   • Q1 = 1.03 (0.63-1.69)

Q1 = <0.5

   • Q2 = 1.00 (referent)

D & B score = 13

Q2 = 0.5

   • Q3 = 0.56 (0.35-0.91)

Q3 = 0.6-0.9

   • Q4 = 0.71 (0.49-1.02)

Q4 = >1.0

Multivariate adjusted HR (95% CI) by duration of walking PA, women

Outcome Measure: Death from ischemic stroke

   • Q1 = 1.38 (0.82-2.33)

   • Q2 = 1.00 (referent)

Cox proportional HR

   • Q3 = 0.56 (0.32-0.97)

   • Q4 = 0.73 (0.48-1.13)

Multivariate adjusted HR (95% CI) by sport PA, men

   • Q1 = 1.34 (0.86-2.08)

   • Q2 = 1.00 (referent)

   • Q3 = 1.22 (0.66-2.25)

   • Q4 = 0.84 (0.45-1.57)

Multivariate adjusted HR (95% CI) by sport PA, women

   • Q1 = 1.07 (0.64-1.77)

   • Q2 = 1.00 (referent)

   • Q3 = 0.62 (0.25-1.58)

   • Q4 = 0.73 (0.31-1.70)

Paganini-Hill and Barreto 2001 [265]

To identify risk factors and preventative measures for stroke in elderly men and women.

• n = 4,722 men, 8,532 women

Baseline survey in 1981- 1982.

Multivariate adjusted RR (95% CI) for total hemorrhagic occlusion by exercise, men

Emphasized role of lifestyle modification in the primary prevention of stroke.

• Sex: Men and women

PA assessment: Questionnaire on amount of hours per day of exercise

   • Q1 = 1.00 (referent)

USA

Age: 44-101 yr

   • Q2 = 0.88

• Characteristics: no previous history of stroke. Residence of a retirement community in Southern California

Q3 = 0.83

Prospective cohort

G1 = <0.5

G2 = <0.1

Multivariate adjusted RR (95% CI) for total hemorrhagic occlusion by exercise, women

G3 = 1+

D & B score = 13

Outcome Measure: Incidence of hemorrhagic occlusion strokes up until December 31, 1998.

   • Q1 = 1.00 (referent)

   • Q2 = 0.91

   • Q3 = 0.85

Poisson Regression 40 year follow up

Pitsavos et al 2004 [266]

To investigate the interaction between PA in men with LVH on stroke mortality.

• n = 489

Number of cases: 67

PA reduced the risk of stroke in men without LVH.

• Sex: Men

PA assessment: Questionnaire

RR (95% CI)

USA

• Age: 40-59 yr

   • G1 = 1.00 (referent)

• Characteristics: Those without LVH

G1 = Sedentary

   • G2 = 0.64 (0.45-0.91)

Prospective cohort

G2 = Moderate

   • G3 = 0.72 (0.51-1.02)

• Corfu Cohort (Greece) from Seven Countries Study

G3 = Hard

D & B score = 12

Outcome Measure: Stroke mortality

Cox proportional HR

Sacco et al 1998 [267]

To investigate the association between LTPA and ischemic stroke.

• n = 369 case, 678 control

Case Subjects were recruited during hospitalization, self referral or from monitoring non hospitalized stroke. Controls were eligible if they had never been diagnosed with stroke and were >39 years.

LTPA was related to a decreased occurrence of ischemic stroke in elderly, multiethnic, urban subjects.

• Sex: Men and women

O R (95% CI) for duration of LTPA and stroke

USA

• Age: > 39 yr

   • G1 = 1.00 (referent)

Case control

• Characteristics: Case Subjects: Diagnosed with first cerebral infarction after July 1, 1993. Control Subjects: Never diagnosed with stroke

   • G2 = 0.42

   • G3 = 0.35

D & B score = 14

   • G4 = 0.31

PA assessment:

Questionnaire

Divided into duration of LTPA (h/wk)

• Northern Manhattan Stroke Study

G1 = 0

G2 = <2

G3 = 2-<5

G4 = ≥ 5

Multivariate conditional logistic regression Baseline data collection from 1982-1983 in East Boston (MA), New Haven (CT) and Iowa and Washington counties (IA).

Simonsick et al 1993 [268]

To examine the association between recreational PA among physically capable older adults and incidence of selected chronic diseases and mortality over 3 and 6 years.

• n = 1,815

After 3 years Iowa

No consistent relationship between PA and stroke was found after 3 or 6 years across all 3 population cohorts.

• Sex: Men and women

• Age: ≥ 65 yrs

OR (95% CI) Stroke and activity level

USA

• Characteristics: Physically capable to do heavy work around the house, walk up and down a flight of stairs and walk a half mile without help.

   • T1 = 0.22 (0.08-0.61)

   • T2 = 1.05 (0.60-1.84)

Prospective cohort

   • T3 = 1.00 (Referent)

PA assessment: Questionnaire

New Haven

D & B score = 12

T1 = High

OR (95% CI) Stroke and activity level

T2 = Moderate and

   • T1 = 1.06 (0.38-2.95)

T3 = Inactive

   • T2 = 1.26 (0.54-2.92)

• Established Populations for Epidemiologic Studies of the Elderly

   • T3 = 1.00 (Referent)

Outcome Measure: Stroke incidence during 3 and 6 year follow-ups.

East Boston

OR (95% CI) Stroke and activity level

   • T1 = 0.59 (0.17-1.95)

Logistic Regression

   • T2 = 1.08 (0.52-2.27)

   • T3 = 1.00 (Referent)

After 6 years

Iowa

OR (95% CI) Stroke and activity level

   • T1 = 0.56 (0.31-1.00)

   • T2 = 0.97 (0.64-1.48)

   • T3 = 1.00 (Referent)

New Haven

OR (95% CI) Stroke and activity level

   • T1 = 1.05 (0.52-2.12)

   • T2 = 1.29 (0.72-2.32)

   • T3 = 1.00 (Referent)

East Boston

OR (95% CI) Stroke and activity level

   • T1 = 1.21 (0.56-2.61)

   • T2 = 1.73 (0.98-3.06)

   • T3 = 1.00 (Referent)

Thrift et al 2002 [269]

To examine whether intracerebral hemorrhage is associated with dynamic or static exercise.

• n = 662

PA assessment: Interview, divided into 3 groups: frequency of vigorous activity

Number of Cases: 331

Findings not significant after multivariate analysis.

• Sex: Men and women

• Age: 18-80 yr

Multivariate OR (95% CI) by frequency of VPA

Australia

• Characteristics: Cases: first episode ofintracerebral hemorrhage Controls: Neighbours of cases

G1 = Never

   • G1 = 1.00 (referent)

Case control

G2 = Rarely

   • G2 = 0.68 (0.36-1.27)

G3 = Once or more per month

   • G3 = 0.66 (0.39-1.11)

D & B score = 14

p = 0.094

OPA level

Multivariate OR (95% CI) by OPA level

G1 = Sedentary

   • G1 = 1.00 (referent)

G2 = Light to moderate

   • G2 = 0.94 (0.59-1.48), p = 0.773

G3 = Heavy

   • G3 = 1.18 (0.57-2.46), p = 0.650

Outcome Measure: Intracerebral hemorrhage

Multiple logistic regression