Introduction
Stroke is the leading cause of death and disability worldwide (Lopez et al., Reference Lopez, Mathers, Ezzati, Jamison and Murray2006) despite the knowledge that its risk can be reduced by the appropriate management of modifiable cardiovascular risk factors (Goldstein et al., Reference Goldstein, Adams, Alberts, Appel, Brass, Bushnell, Culebras, DeGraba, Gorelick, Guyton, Hart, Howard, Kelly-Hayes, Nixon and Sacco2006).
Studies show that parental stroke is an independent risk factor for stroke (Jousilahti et al., Reference Jousilahti, Rastenyte, Tuomilehto, Sarti and Vartiainen1997; Polychronopoulos et al., Reference Polychronopoulos, Gioldasis, Ellul, Metallinos, Lekka, Paschalis and Papapetropoulos2002) and, in those who have suffered a primary cerebrovascular event, there is evidence of an association between family history of stroke and modifiable risk (Flossmann and Rothwell, Reference Flossmann and Rothwell2005; Lindgren et al., Reference Lindgren, Lovkvist, Hallstrom, Hoglund, Jonsson, Kristoffersson, Luthman, Petersen and Norrving2005). Lindgren found that first-degree relatives of people who had suffered a stroke reported a higher prevalence of hypertension than did relatives of controls who had not suffered a stroke (Lindgren et al., Reference Lindgren, Lovkvist, Hallstrom, Hoglund, Jonsson, Kristoffersson, Luthman, Petersen and Norrving2005), but this finding was not linked to any validated measurements.
There is evidence to suggest that individuals with a parental history of cardiovascular disease do not pursue self-initiated, sustained change in modifiable risk factors (Kip et al., Reference Kip, McCreath, Roseman, Hulley and Schreiner2002). Thus, opportunities may exist for improving primary prevention of stroke through targeted screening of individuals with a relevant parental history, for risk identification and reduction. However, there is a lack of evidence of the relative prevalence of modifiable risk factors between those with and without a parental history of stroke or of the potential value of adopting such a strategy in general practice.
This study therefore aimed to compare, in the general population, the prevalence of stroke risk factors, among people with a parental history of stroke, to those in a control group of individuals, of similar age, gender and social class, with no parental stroke history.
Methods
Design
We carried out a case–control study of individuals from 11 selected general practices in Northern Ireland (NI) between August 2004 and July 2005. These practices were chosen to represent the general population of NI and included a range of geographical location, practice list size, deprivation scores and population density. Three hundred individuals from the total population in each practice falling between the ages of 40 and 65 years (the ‘invitees’) were selected using random number assignment (Urbaniak and Plous, Reference Urbaniak and Plous2005). A questionnaire was sent relating to a history of parental stroke, sociodemographic information and consent for contact by the researcher. A pre-paid pre-addressed envelope was included for return. Replies received within six weeks of posting (the ‘responders’) were considered for recruitment to the study; those received later were excluded.
Recruitment
Within each practice those with a parental history of stroke were matched to individuals with no parental stroke history. Matching was performed on gender, age (±5 years) and socioeconomic status (±10) using a commonly used regional index of multiple deprivation (IMD; Noble et al., Reference Noble, Smith, Wright, Dibben and Lloyd2001) (scored from 0 to 100) derived from the postal code of the place of residence. Potential participants were telephoned on at least three occasions, at a variety of times including evening, before attempted contact was abandoned. Those contacted were invited to attend their own general practice surgery (the ‘attendees’) to meet with the researcher for an interview and medical examination.
Data collection
At interview, following written consent, biographical and sociodemographic data were collected by the researcher. Data were gathered about the symptoms observed at the time of the stroke of their parent in order to provide some validation of the diagnosis. Participants were asked about their own relevant past medical history, current medications and smoking and alcohol consumption. Height was measured (to nearest 0.1 cm) using a stadiometer (Seca 214 Portable Stadiometer, Hamburg, Germany) and weight (to nearest 1 kg) using mechanical clinical scales (Seca 761 Medical Scales, Hamburg, Germany). Waist circumference (widest point of abdomen below bottom of rib cage and above iliac crest) and hip circumference (maximum width over greater trochanters of the pelvic girdle) were measured using a tape measure to the nearest 1 cm. Body mass index (BMI) (kg/m2) and waist–hip ratio (WHR) were each derived. Resting non-invasive brachial arterial blood pressure was measured (Williams et al., Reference Williams, Poulter, Brown, Davis, McInnes, Potter, Sever and McG Thom2004) twice, with a 3-min gap between the first and second measurements, using an approved validated automated oscillometric device (Omron HEM-705CP, Omron, Schaumburg, IL, USA) (O’Brien et al., Reference O’Brien, Mee, Atkins and Thomas1996; The British Hypertension Society, 2006); the second measurement was used for analysis. Participants with a blood pressure above 140/90 mmHg (Williams et al., Reference Williams, Poulter, Brown, Davis, McInnes, Potter, Sever and McG Thom2004) were asked to return to their own general practitioner for further checks. Results of blood tests were returned to each practice with abnormal results highlighted for attention.
Participants completed one validated questionnaire for dietary habit (Dietary Instrument for Nutrition Education questionnaire (Roe et al., Reference Roe, Strong, Whiteside, Neil and Mant1994; Little and Margetts, Reference Little and Margetts1996)) and another for exercise habit (Baecke Physical Activity Questionnaire; Baecke et al., Reference Baecke, Burema and Frijters1982).
Data analysis
A number of dichotomous variables were derived for analysis. ‘Raised blood pressure’ was defined as a history of hypertension, currently taking anti-hypertensive medication, systolic blood pressure (SBP) above 140 mmHg or diastolic blood pressure (DBP) above 90 mmHg (Williams et al., Reference Williams, Poulter, Brown, Davis, McInnes, Potter, Sever and McG Thom2004). ‘Raised cholesterol’ included participants with TChol/high-density lipoprotein (HDL) ratio above 5.0. An individual was deemed to have diabetes if they reported a history of diabetes, were currently taking diabetic medication or their fasting blood glucose was above 6.9 mmol/L (The Expert Committee on the Diagnosis and Classification of Diabetes Mellitus, 1997). Alcohol intake was categorised using recommeded limits for weekly consumption (21 U for men; 14 U for women; Mulvihill et al., Reference Mulvihill, Taylor and Waller2005). Using a conversion of 8 g alcohol per unit (Miller et al., Reference Miller, Heather and Hall1991), units consumed per week were converted into grams of alcohol per day and recoded into another categorical variable (greater or less than 60 g of alcohol per day), based on a meta-analysis detailing alcohol levels that were clinically significant (Reynolds et al., Reference Reynolds, Lewis, Nolen, Kinney, Sathya and He2003). ‘Metabolic syndrome’ was defined by the presence of central obesity (abdominal circumference ⩾94 cm for men and ⩾80 cm for women) plus two of the following: raised triglycerides (>1.7 mmol/L), reduced HDL cholesterol (<1.03 mmol/L for males and 1.29 mmol/L for females), raised blood pressure (SBP ⩾ 130 mmHg or DBP ⩾ 85 mmHg)) and raised fasting blood glucose (⩾5.6 mmol/L; Alberti et al., Reference Alberti, Zimmet and Shaw2005).
SPSS (Statistics Computer Programme 15.0.1 – Release November 2006) was used for data entry and analysis. Descriptive statistics were used to summarise data. Continuous variables were compared for case– control pairs using a t-test for paired samples. Categorical variables were analysed using McNemar’s test for non-parametric data. For variables with missing data, case–control data pairs were removed from the analysis. Multiple regression analysis was used to adjust paired comparisons of blood pressure between cases and controls for confounding variables.
Sample size was based on an assumption of risk factor prevalence (eg, current cigarette smoking or hypertension) of 30% in the control group. A 50% difference between the groups (ie, 30% versus 45%) in risk factor prevalence would be detected at the 5% significance level and power greater than 80% with independent samples of 200 cases and 200 controls.
Ethics approval
Ethics approval to conduct the study was sought and granted by the Queen’s University Belfast Research Ethics Committee.
Results
The total population of the 11 practices from which the study sample was drawn was 104 207, which represents 6.2% of the NI population. Of the 3300 questionnaires sent, 1061 (32%) were returned within six weeks of posting. Completed data were obtained for 199 case–control pairs (398 individuals; Figure 1). Comparing the study attendees to invitees, attendees were slightly older (mean age 53 years (SD 6.7) versus 51.2 years (SD 7.0; P < 0.001)), had lower levels of deprivation (IMD 16.7 (SD 16.4) versus 23.2 (SD 20.8; P < 0.001)) and had a greater representation of females (54.5% versus 51.4%; Table 1). Comparison of cases and controls in respect of mean age (53.1 (SD 6.7) years versus 53.0 (SD 6.7) years) and mean deprivation status (IMD 16.8 (SD 16.7) versus 16.6 (SD 16.2)) confirmed successful matching on those characteristics.
Flow diagram of recruitment process
Summary characteristics of ‘invitees’, ‘responders’ and ‘attendees’
Mean diastolic BP (DBP) was 2.7 mmHg higher and mean systolic BP (SBP) 5.7 mmHg higher in cases than in controls (Table 2). More cases than controls had ‘raised blood pressure’, but the difference was not statistically significant (P = 0.07). Cases were more likely than controls to have a prior diagnosis of hypertension and to be on anti-hypertensive medication.
Summary of characteristics of 199 case–control pairs (continuous data)
BMI = body mass index; DINE = Dietary Instrument for Nutrition Education.
A greater proportion of cases than controls declared a history of hypercholesterolaemia, diabetes mellitus, previous stroke, previous myocardial infarction, atrial fibrillation and ischaemic heart disease, but these differences were not statistically significant (Table 3).
Summary of characteristics of 199 case–control pairs (categorical data)
aInsufficient numbers to permit formal testing.
Similar proportions of cases and controls were both current smokers and considered to have a ‘smoking-associated risk’ (current smoker or exsmoker of <10 years). Cases consumed a mean of 3.7 U of alcohol more per week than their paired controls. In addition, significantly more cases than controls consumed above the recommended limits. No significant differences were found between cases and controls for measures of lipids, glucose, diabetes, BMI, WHR, dietary habit or physical activity. However, it was noted that differences in fasting blood glucose and triglycerides between groups approached statistical significance.
A regression analysis was carried out to adjust the blood pressure comparisons for BMI, smoking, alcohol consumption, fasting blood glucose and diabetic mellitus history (Table 4). After adjustment, the difference between cases and controls remained statistically significant for SBP, but for DBP the difference between cases and controls failed to reach significance.
Multiple regression analysis of mean differences between cases and controls in paired measures of blood pressure
BMI = body mass index.
alog10 scale (after addition of 1 unit/week for alcohol).
Discussion
This study shows that screening individuals with a parental history of stroke for stroke risk factors returns significantly higher mean systolic and DBP and higher alcohol consumption compared with those with no parental history of stroke. Furthermore, the relative prevalence of hypertension and of alcohol consumption above recommended limits are higher among those with a parental history of stroke.
After regression analysis, with adjustment for other risk factors, a relationship between higher mean SBP and positive parental history of stroke remains. The mean difference between age- and gender-matched pairs in systolic BP is 5.7 mmHg, with a mean BP of 146/88 mmHg for those with a parental history of stroke. Stroke risk has been shown to have a direct and linear relationship with blood pressure down to at least 115/75 mmHg (Lewington et al., Reference Lewington, Clarke, Qizilbash, Peto and Collins2002), and a meta-analysis has shown that a reduction of 10 mmHg of systolic BP in those under 60 years was associated with a 40% to 50% reduction in stroke risk (Lawes et al., Reference Lawes, Bennett, Feigin and Rodgers2004).
The findings of our study add to the evidence from previous work of the relevance of self-reported hypertension in those with a family history of stroke (Flossmann and Rothwell, Reference Flossmann and Rothwell2005; Lindgren et al., Reference Lindgren, Lovkvist, Hallstrom, Hoglund, Jonsson, Kristoffersson, Luthman, Petersen and Norrving2005). In keeping with the relative importance of blood pressure for all stroke risk it may be that blood pressure is the most important element in the manifestation of that heritability. Either way the findings lend weight to the argument for targeting the offspring of patients with a history of stroke for blood pressure screening and monitoring in routine clinical practice.
The higher mean weekly alcohol consumption among cases compared with controls (14 versus 10 U/week; P < 0.01) and greater proportion drinking above recommended limits (29% versus 20%; P < 0.05) were unexpected findings. These patterns of alcohol consumption may reflect other family-influenced lifestyle behaviours. However, their clinical significance is unclear given the lack of consensus about the role that light-to-moderate alcohol consumption plays in stroke risk (Mukamal et al., Reference Mukamal, Chung, Jenny, Kuller, Longstreth, Mittleman, Burke, Cushman, Beauchamp and Siscovick2005). A greater consensus exists around the role that alcohol consumed in large quantities plays in stroke risk, with a relative risk of 1.69 for total stroke for those consuming more than 60 g of alcohol per day compared with abstainers (Reynolds et al., Reference Reynolds, Lewis, Nolen, Kinney, Sathya and He2003). Although not statistically significant, we observed a trend towards a higher proportion of people with a parental history of stroke consuming more than 60 g alcohol per day. The lack of statistical significance may have been due to an inadequate sample size given that the prevalence of this level of alcohol consumption was small (4%, 1%) in both groups.
Strengths
This study was based in general practice, the setting in which application of its results is likely to be most relevant. The method used an introductory letter from the practice where the study was to take place. It was considered that this added credibility to the researcher’s letter of invitation and contributed to the willingness of those contacted to participate, thus optimising the potential to obtain a representative sample.
A single researcher carried out all the assessments, thus providing standardisation of measurements and consistency of approach. Conversely, the potential for bias in measurement from an observer who was not blind to the group status of participants should also be acknowledged. These criticisms are at least partially addressed in that the study used, where possible, validated research tools such as the diet and exercise questionnaires and an automatic oscillometric sphygmomanometer, which were not observer-dependent. Biochemical analysis of blood samples was done blind to group status.
Limitations
Although we observed differences in measured BP, we also found that more of those with than without a parental history of stroke reported a history of hypertension (33% versus 20%; P < 0.01) and were on anti-hypertensive treatment. Possible explanations for this are that those with a positive family history were aware of blood pressure as a stroke risk factor and pro-actively sought BP measurement or that physicians’ awareness of parental history of stroke prompted BP screening. The observation prompts speculation as to whether treatment may have masked a greater true mean difference in blood pressure between cases and controls.
Within the study protocol there was no reliable way of confirming the parental history of stroke or stroke subtype as in many cases the parent was dead and their medical notes would not be accessible. However, only those who were able to recall signs and symptoms in keeping with the FAS test (Nor et al., Reference Nor, McAllister, Louw, Dyker, Davis, Jenkinson and Ford2004) (facial, arm or leg weakness or a disorder of speech) that suggested their parent had suffered a stroke were included in the study. Reassuringly, from work done in the Family Heart Study population, the κ-statistics for agreement between proband-reported family history and the self-reported personal history of stroke in members of the proband’s family were 0.77 for a proband-reported father’s history versus father’s self-reported history of stroke and 0.76 for a proband-reported mother’s history versus mother’s self-reported history (Kornegay et al., Reference Kornegay, Liao, Bensen, Province, Folsom and Ellison1997). Although the method adopted is unvalidated and lacks diagnostic rigour, it represents real-world clinical practice in which practitioners often do not have access to the notes of their patients’ parents.
There may have been an attendance bias within the study as attendees were slighty older, more likely to be female and less socioeconomically deprived than invitees and responders. It may be that those attending were more health-conscious than non-responders and non-attendees. This may explain the failure to observe any difference in smoking prevalence between cases and controls: both may already have been following good advice in respect of stroke prevention and healthy lifestyles. However, the overall reported prevalence of smoking among the study population (22%) is similar to recent estimates for smoking habits in Great Britain (General Household Survey, 2004), suggesting that the population studied is representative of the wider population in respect of smoking habits.
Generalisability of the findings
Within each general practice a random sample was drawn to avoid selection bias. However, a non-response bias, with consequent over-representation within our sample of those who were more health conscious, slightly older and of higher socioeconomic status, may remain a possibility as acknowledged above.
Conclusion
The results of this study suggest that targeting the offspring of patients with stroke as part of a screening strategy in respect of blood pressure may be worthwhile in identifying those for whom primary prevention would be a worthwhile step towards reducing the stroke burden. Given the assertion that individuals with a parental history of cardiovascular disease do not pursue self-initiated, sustained change in modifiable risk factors (Kip et al., Reference Kip, McCreath, Roseman, Hulley and Schreiner2002) and the relative importance of blood pressure reduction for the primary prevention of stroke, such a strategy could be justified (Goldstein et al., Reference Goldstein, Adams, Alberts, Appel, Brass, Bushnell, Culebras, DeGraba, Gorelick, Guyton, Hart, Howard, Kelly-Hayes, Nixon and Sacco2006).
However, with the exception of alcohol consumption, the results do not support the adoption of a strategy to screen these individuals for all other commonly cited stroke risk factors. Further work would be required for a full understanding of the implications of the findings in respect of consumption of alcohol.
The introduction of a screening strategy to assess these risk factors in the offspring of all patients with a history of stroke would have implications for service provision. To avoid redundancy of effort, to ensure the appropriate use of limited resources and to provide a definitive answer of likely benefit, the adoption of a targeted screening strategy for blood pressure among the offspring of stroke patients would need to be measured against population strategies currently in use (NHS Confederation (Great Britain), 2003; British Cardiac Society et al., 2005; Goldstein et al., Reference Goldstein, Adams, Alberts, Appel, Brass, Bushnell, Culebras, DeGraba, Gorelick, Guyton, Hart, Howard, Kelly-Hayes, Nixon and Sacco2006; Jessani et al., Reference Jessani, Watson, Cappuccio and Lip2006).
Acknowledgements
The study was funded by the Research and Development Office, Northern Ireland through a United Kingdom, National Health Service, Primary Care Researcher Development Award, awarded to Dr Nigel Hart. The authors declare no conflict of interest.
