Subjects and methods

The study was in east Lancashire, a health district including the towns of Blackburn and Burnley, with a total population of 534 287 in the 1995 general practice register. Stroke cases were identified from a district wide, population based stroke register between 1 July 1994 and 30 June 1995 with 103 of 118 general practices participating. Cases from a further 10 (8%) practices were inconsistently notified or verified, so the total practice inclusion rate was 79% (93 practices). These practices served a population of 405 272, of whom 388 821 were under 80 years of age (161 978 aged 40-79 years), the denominator for the study.

A case-control design was used, restricted to patients aged under 80 with their first ever strokes because strokes below this age can be regarded as preventable.14 15 The main hypothesis was that there was a difference in risk of stroke between hypertensive patients who were treated but not well controlled and those well controlled as well as subjects who were non-hypertensive. We used an arbitrary definition of hypertension (!160/95 mm Hg)16 likely to be in routine use before the study and assumed the prevalence of treated hypertensive patients who were not well controlled (!150/90 mm Hg) to be 14% of the population aged under 80. Applying these variables showed that 220 cases and 440 controls were required to detect an odds ratio of 2 compared with non-hypertensive subjects with 80% power when using a two sided α level of 5%. For each case two controls matched with the case for sex and age (within two years) and without a history of stroke were selected from the same practice register.

Read more  Is Ginger a Safe and Effective Treatment for Nausea?

All blood pressure readings with the dates and drug treatment before the index date of stroke were recorded from the notes or the practice computer. The classification of who met definitions of “hypertension” was conducted on computer blind to case or control status. Each set of notes was also systematically searched for other known prespecified risk factors.

Hypertension was defined as above blood pressure levels on two or more occasions within any three month period or a history of antihypertensive treatment at any time. Baseline blood pressure was either the average of these two readings if they met the criteria for hypertension or the blood pressure immediately before treatment first started if the criteria had not been met at that time. The quality of hypertension control was assessed by using the mean blood pressure recorded in the last one, three, and five years before the index date. Treated hypertension was defined from the documented use (prescription) of drugs appropriate for hypertension.

Analysis–Matched case-control analysis was by conditional logistic regression,17 odds ratios being used to determine associations. Population attributable risk was calculated as the difference between overall risk in the population and the population risk that might be achieved if blood pressure were maintained at the “well controlled” level, divided by overall risk in the population. This was calculated in the population aged over 40 years because strokes are rare in people under 40.