Family History of Premature Cardiovascular Disease
Family History of Premature Cardiovascular Disease
Glasgow is the largest city in Scotland with a population of 2.3 million representing 41% of Scotland's population. The Greater Glasgow area is a contiguous urban area within the city of Glasgow with a population of 1.2 million. The Glasgow Blood Pressure Clinic (GBPC) located in the Greater Glasgow area is the largest and the main specialist hypertension clinic in Glasgow and provides secondary and tertiary level service to patients with hypertension in Glasgow. Patients are referred to the GBPC if their blood pressures (BPs) are not controlled in primary care with at least three drugs or if there is evidence of high-risk factors such as early-onset hypertension, features of secondary hypertension, or family history or premature CVD. Structured instruments are used to collect data from all patients attending the clinic and are stored electronically in a single computerized database, which contains information on 16 011 patients attending the clinic from 1969 until 2011. The West of Scotland research ethics service (WoSRES) of the National Health Service has approved the study of the GBPC database (11/WS/0083).
The GBPC employs specialist hypertension nurses who are experienced and highly trained in BP measurement. The procedure required subjects to rest for 5 min in a seated position before BP was manually measured using standard sphygmomanometers and Korotkoff sounds (phase V) were used for estimating diastolic BP (DBP). Three BP measurements were performed, 1min apart, and the mean of the second and third measurements was recorded. Patients attending the clinic were advised to take their regular medications as usual. Each patient attended the same clinic; therefore at each visit, their BP measurement would occur in the same 3 h time window either in the morning or afternoon. Height and weight of all patients were measured using standardized equipment during each visit in order to calculate body mass index (BMI). Blood samples were collected at baseline and at regular intervals for estimation of routine haematological and biochemical indices. All biochemical investigations were performed at the Western Infirmary clinical laboratory service. Glomerular filtration rate (GFR) was estimated using modification of diet in renal disease equation. Both tobacco (any vs. none) and alcohol use (quantity and frequency of consumption) were assessed using a structured format during the clinic visit.
Records of patients who attended the GBPC from 1969 to 2011 were extracted from the database and reviewed. Each patient attending the clinic completed a structured questionnaire on health details of first-degree relatives (parents and siblings): alive/dead, number of full brothers and full sisters, history of hypertension, myocardial infarction and stroke, age at death, age at heart attack/stroke, and age at diagnosis of hypertension. The cause of death for each parent or sibling was recorded as free text and this was manually classified into eight categories—respiratory, possible respiratory, CV, possible CV, cancer, possible cancer, other, and not known. Family history of CVD was determined using the ASSIGN criteria for premature vascular disease—development of heart disease or stroke before the age of 60 years in parent or sibling. We divided patients into four groups based on their history of premature CVD in their parents or first-degree relatives: P-FH and P-FH were, respectively, groups with or without a parental history (either or both parents) of premature CVD. FDR-FH and FDR-FH were groups with or without a first-degree relative with premature CVD.
Records kept by the General Register Office for Scotland ensured notification of a subject's death (provided that it occurred in the UK) together with the primary cause of death according to the International Classification of Diseases, 10th Revision, Version for 2007 (ICD-10), codes. We considered CV deaths (CV mortality; ICD-10 codes I00-I99), ischaemic heart disease deaths (IHD mortality; ICD-10 codes 120-I25), and stroke deaths (stroke mortality; ICD-10 codes I60-I69) in the analysis. Deaths not due to these conditions were classified as non-CV deaths. Mortality data were collected up to April 2011.
Data on refilled prescriptions were available on a subset of patients through the Information Services Division (ISD) for NHS Scotland. The Prescribing Information System holds information on 100% of NHS Scotland prescriptions dispensed within the community and claimed for payment by a pharmacy contractor (i.e. pharmacy, dispensing doctor, or appliance supplier). It does not include data on prescriptions dispensed but not claimed (likely to be very small) or prescriptions prescribed but not submitted for dispensing by a patient. Prescription data were available from 2004 onwards on alive subjects. Although the subgroup in whom prescription data were available may not be representative, we selected a group in whom we had prescription data for at least 5 years to assess differences in adherence to CV drugs based on family history. We calculated the average annual refill rate of antihypertensive, antiplatelet, and lipid-lowering drugs for each patient during this period of review as measures of adherence. This does not account for drugs dispensed to the patient but subsequently not taken by the patient in accordance with dosage instructions.
Differences in baseline characteristics in groups with and without positive family history of CVD were explored using independent t-test (for continuous variables) and χ test (for categorical variables).
We explored the effect of positive family history on longitudinal changes in BP, cholesterol, BMI, and estimated GFR (eGFR) using generalized estimating equations (GEE). Individuals with at least four annual measurements in the first 5 years of follow-up, a 5-year minimum follow-up period, and survival up to a minimum of 5 years were included in this analysis. The association was adjusted for conventional covariates—baseline age, sex, alcohol and tobacco use, BMI, cholesterol, and eGFR as appropriate. All non-missing pairs of data were used in estimating the working correlation parameters.
The Cox proportional hazard (Cox-PH) models were used to analyse the relationship between a positive family history of CVD on all-cause, CV, IHD, stroke, and non-CV mortality. We present the unadjusted (Model 1) and adjusted results separately (Models 2–4). A variable on year of first visit strata (epochs) was used to adjust the secular trend in mortality (Model 2) and was divided into five categories (first visit 1977 or before, between years 1978–85, 1986–93, 1994–2001, 2002 and thereafter). Model 3 is adjusted for baseline age, sex, and epochs. Model 4 was adjusted for baseline age, gender, epochs, BMI, smoking status (never vs. ever), systolic BP (SBP), alcohol use, baseline prevalence of CVD, and chronic kidney disease (CKD) status (eGFR <60, eGFR ≥60). The proportional hazards assumption was verified through examination of log-minus-log plots.
Propensity score matching (nearest neighbour) was used to match patients with and without a positive family history of CVD. Cox-PH models were generated in the matched population to study the independent association of family history of CVD and different types of mortality outcomes.
The discriminatory power of positive family history of CVD in predicting CV mortality was assessed using 'C-statistics', net-reclassification improvement, and integrated discrimination improvement. Stata Version 12.0 (Statacorp) was used for all statistical analysis.
Methods
Study Setting and Study Population
Glasgow is the largest city in Scotland with a population of 2.3 million representing 41% of Scotland's population. The Greater Glasgow area is a contiguous urban area within the city of Glasgow with a population of 1.2 million. The Glasgow Blood Pressure Clinic (GBPC) located in the Greater Glasgow area is the largest and the main specialist hypertension clinic in Glasgow and provides secondary and tertiary level service to patients with hypertension in Glasgow. Patients are referred to the GBPC if their blood pressures (BPs) are not controlled in primary care with at least three drugs or if there is evidence of high-risk factors such as early-onset hypertension, features of secondary hypertension, or family history or premature CVD. Structured instruments are used to collect data from all patients attending the clinic and are stored electronically in a single computerized database, which contains information on 16 011 patients attending the clinic from 1969 until 2011. The West of Scotland research ethics service (WoSRES) of the National Health Service has approved the study of the GBPC database (11/WS/0083).
Clinical Measurements
The GBPC employs specialist hypertension nurses who are experienced and highly trained in BP measurement. The procedure required subjects to rest for 5 min in a seated position before BP was manually measured using standard sphygmomanometers and Korotkoff sounds (phase V) were used for estimating diastolic BP (DBP). Three BP measurements were performed, 1min apart, and the mean of the second and third measurements was recorded. Patients attending the clinic were advised to take their regular medications as usual. Each patient attended the same clinic; therefore at each visit, their BP measurement would occur in the same 3 h time window either in the morning or afternoon. Height and weight of all patients were measured using standardized equipment during each visit in order to calculate body mass index (BMI). Blood samples were collected at baseline and at regular intervals for estimation of routine haematological and biochemical indices. All biochemical investigations were performed at the Western Infirmary clinical laboratory service. Glomerular filtration rate (GFR) was estimated using modification of diet in renal disease equation. Both tobacco (any vs. none) and alcohol use (quantity and frequency of consumption) were assessed using a structured format during the clinic visit.
Family History Assessment
Records of patients who attended the GBPC from 1969 to 2011 were extracted from the database and reviewed. Each patient attending the clinic completed a structured questionnaire on health details of first-degree relatives (parents and siblings): alive/dead, number of full brothers and full sisters, history of hypertension, myocardial infarction and stroke, age at death, age at heart attack/stroke, and age at diagnosis of hypertension. The cause of death for each parent or sibling was recorded as free text and this was manually classified into eight categories—respiratory, possible respiratory, CV, possible CV, cancer, possible cancer, other, and not known. Family history of CVD was determined using the ASSIGN criteria for premature vascular disease—development of heart disease or stroke before the age of 60 years in parent or sibling. We divided patients into four groups based on their history of premature CVD in their parents or first-degree relatives: P-FH and P-FH were, respectively, groups with or without a parental history (either or both parents) of premature CVD. FDR-FH and FDR-FH were groups with or without a first-degree relative with premature CVD.
Outcome Assessment
Records kept by the General Register Office for Scotland ensured notification of a subject's death (provided that it occurred in the UK) together with the primary cause of death according to the International Classification of Diseases, 10th Revision, Version for 2007 (ICD-10), codes. We considered CV deaths (CV mortality; ICD-10 codes I00-I99), ischaemic heart disease deaths (IHD mortality; ICD-10 codes 120-I25), and stroke deaths (stroke mortality; ICD-10 codes I60-I69) in the analysis. Deaths not due to these conditions were classified as non-CV deaths. Mortality data were collected up to April 2011.
Adherence Assessment
Data on refilled prescriptions were available on a subset of patients through the Information Services Division (ISD) for NHS Scotland. The Prescribing Information System holds information on 100% of NHS Scotland prescriptions dispensed within the community and claimed for payment by a pharmacy contractor (i.e. pharmacy, dispensing doctor, or appliance supplier). It does not include data on prescriptions dispensed but not claimed (likely to be very small) or prescriptions prescribed but not submitted for dispensing by a patient. Prescription data were available from 2004 onwards on alive subjects. Although the subgroup in whom prescription data were available may not be representative, we selected a group in whom we had prescription data for at least 5 years to assess differences in adherence to CV drugs based on family history. We calculated the average annual refill rate of antihypertensive, antiplatelet, and lipid-lowering drugs for each patient during this period of review as measures of adherence. This does not account for drugs dispensed to the patient but subsequently not taken by the patient in accordance with dosage instructions.
Statistical Analysis
Differences in baseline characteristics in groups with and without positive family history of CVD were explored using independent t-test (for continuous variables) and χ test (for categorical variables).
We explored the effect of positive family history on longitudinal changes in BP, cholesterol, BMI, and estimated GFR (eGFR) using generalized estimating equations (GEE). Individuals with at least four annual measurements in the first 5 years of follow-up, a 5-year minimum follow-up period, and survival up to a minimum of 5 years were included in this analysis. The association was adjusted for conventional covariates—baseline age, sex, alcohol and tobacco use, BMI, cholesterol, and eGFR as appropriate. All non-missing pairs of data were used in estimating the working correlation parameters.
The Cox proportional hazard (Cox-PH) models were used to analyse the relationship between a positive family history of CVD on all-cause, CV, IHD, stroke, and non-CV mortality. We present the unadjusted (Model 1) and adjusted results separately (Models 2–4). A variable on year of first visit strata (epochs) was used to adjust the secular trend in mortality (Model 2) and was divided into five categories (first visit 1977 or before, between years 1978–85, 1986–93, 1994–2001, 2002 and thereafter). Model 3 is adjusted for baseline age, sex, and epochs. Model 4 was adjusted for baseline age, gender, epochs, BMI, smoking status (never vs. ever), systolic BP (SBP), alcohol use, baseline prevalence of CVD, and chronic kidney disease (CKD) status (eGFR <60, eGFR ≥60). The proportional hazards assumption was verified through examination of log-minus-log plots.
Propensity score matching (nearest neighbour) was used to match patients with and without a positive family history of CVD. Cox-PH models were generated in the matched population to study the independent association of family history of CVD and different types of mortality outcomes.
The discriminatory power of positive family history of CVD in predicting CV mortality was assessed using 'C-statistics', net-reclassification improvement, and integrated discrimination improvement. Stata Version 12.0 (Statacorp) was used for all statistical analysis.