Using the Bland-Altman method to measure agreement with repeated measures Medical researchers often need to compare two methods of measurement, or a new method with an established one, to determine whether these two methods can be used interchangeably or the new method can replace the established one. 1 -6 In most of these situations, the 'true' value of the measured quantity is unknown.In a series of articles, Bland and Altman 7 -9 advocated the use of a graphical method to plot the difference scores of two measurements against the mean for each subject and argued that if the new method agrees sufficiently well with the old, the old may be replaced. Here the idea of agreement plays a crucial role in method comparison studies. There are numerous published clinical and laboratory studies evaluating agreement between two measurement methods using Bland -Altman analysis. The original Bland -Altman publication 7 has been cited on more than 11 500 occasions-compelling evidence of its importance in medical research.The Bland -Altman method calculates the mean difference between two methods of measurement (the 'bias'), and 95% limits of agreement as the mean difference (2 SD) [or more precisely (1.96 SD)]. It is expected that the 95% limits include 95% of differences between the two measurement methods. The plot is commonly called a Bland -Altman plot and the associated method is usually called the Bland -Altman method. The Bland -Altman method can even include estimation of confidence intervals for the bias and limits of agreement, but these are often omitted in research papers. 8 The presentation of the 95% limits of agreement is for visual judgement of how well two methods of measurement agree. The smaller the range between these two limits the better the agreement is. The question of how small is small depends on the clinical context: would a difference between measurement methods as extreme as that described by the 95% limits of agreement meaningfully affect the interpretation of the results?Repeated measurements for each subject are often used in clinical research. Two recent articles in the British

The generalized estimating equation (GEE) approach is a widely used statistical method in the analysis of longitudinal data in clinical and epidemiological studies. It is an extension of the generalized linear model (GLM) method to correlated data such that valid standard errors of the parameter estimates can be drawn. Unlike the GLM method, which is based on the maximum likelihood theory for independent observations, the GEE method is based on the quasilikelihood theory and no assumption is made about the distribution of response observations. Therefore, Akaike's information criterion, a widely used method for model selection in GLM, is not applicable to GEE directly. However, Pan (Biometrics 2001; 57: 120-125) proposed a model-selection method for GEE and termed it quasilikelihood under the independence model criterion. This criterion can also be used to select the best-working correlation structure. From Pan's methods, I developed a general Stata program, qic, that accommodates all the distribution and link functions and correlation structures available in Stata version 9. In this paper, I introduce this program and demonstrate how to use it to select the best working correlation structure and the best subset of covariates through two examples in longitudinal studies.

Abstract-The linkage and association between inherent blood pressure and underlying genotype is potentially confounded by antihypertensive treatment. We estimated blood pressure variance components (genetic, shared environmental, individual-specific) in 767 adult volunteer families by using a variety of approaches to adjusting blood pressure of the 244 subjects (8.2%) receiving antihypertensive medications. The additive genetic component of variance for systolic pressure was 73.9 mm Hg 2 (SE, 8.8) when measured pressures (adjusted for age by gender within each generation) were used but fell to 61.4 mm Hg 2 (SE, 8.0) when treated subjects were excluded. When the relevant 95th percentile values were substituted for treated systolic pressures, the additive genetic component was 81.9 mm Hg 2 (SE, 9.5), but individual adjustments in systolic pressure ranged from Ϫ53.5 mm Hg to ϩ64.5 mm Hg (mean, ϩ17.2 mm Hg). Instead, when 10 mm Hg was added to treated systolic pressure, the additive genetic component rose to 86.6 mm Hg 2 (SE, 10.1). Similar changes were seen in the shared environment component of variance for systolic pressure and for the combined genetic and shared environmental (ie, familial) components of diastolic pressure. There was little change in the individual-specific variance component across any of the methods. Therefore, treated subjects contribute important information to the familial components of blood pressure variance. This information is lost if treated subjects are excluded and obscured by treatment effects if unadjusted measured pressures are used. Adding back an appropriate increment of pressure restores familial components, more closely reflects the pretreatment values, and should increase the power of genomic linkage and linkage disequilibrium analyses. Key Words: antihypertensive therapy Ⅲ blood pressure Ⅲ genetics Ⅲ human Ⅲ epidemiology F amily studies can be used to estimate the relative contribution of genes and environment to blood pressure (BP) variation 1,2 and are the substrate for genomic discovery based on linkage and association analyses. [3][4][5][6][7][8][9] The key to successful genetic discovery is to use phenotypes that reflect the underlying genotype as closely as possible. BP is usually adjusted by regression methods for covariates such as age and gender. The usual regression techniques used to adjust for covariates are inappropriate for adjusting BP for the effects of antihypertensive treatment because they result in treated levels having an average of zero residuals rather than the extreme residuals they deserve, given their pretreatment pressures. Yet, no standardized approach for dealing with pressure measurements for treated individuals has emerged.Among genome scans, some use measured pressure, 3 but more commonly, treated individuals are excluded. 4 -7 Underlying hypertension is sometimes presumed and hypertensive values substituted for treated pressures. 7,8 Observed blood pressure ranking is lost in these approaches, and adjustments do not necessarily reflect possi...

No abstract

Abstract-The correlations between systolic blood pressure (SBP) and diastolic blood pressure (DBP), and between SBP and body mass index (BMI), might result from genetic or environmental factors that determine variation in 2 or more phenotypes and are shared by family members. In 767 adult nuclear families (nϭ2912 individuals, including 66 pairs of monozygotic twins and 84 pairs of dizygotic twins), we used a multivariate normal model and the software FISHER to estimate genetic and environmental components of variation and covariation. Mean phenotypes were adjusted for age, gender, and generation, and for antihypertensive treatment. Genetic and shared family environmental factors accounted for 46% and 31% of total variance in SBP, respectively. Adjustment of SBP for DBP reduced considerably both the additive genetic (86.7 to 21.0) and shared environmental (59.7 to 21.0) components of variance. Smaller reductions in genetic (86.7 to 84.9) and shared environmental (59.7 to 51.1) components were observed after adjustment of SBP for BMI. For SBP and DBP, the correlation between the effects of genes was 1.00 and between shared environmental effects was 0.52. For SBP and BMI the correlations were 0.30 for genetic and 0.22 for shared environmental effects. Our findings suggest that the same genes and many of the same family environmental factors determine variation in both SBP and DBP. In contrast, SBP and BMI share genetic and family environmental determinants to a lesser degree. These observations are relevant to multifactorial cardiovascular risk reduction based on genetic and family environmental approaches. (Hypertension. 2002;40:7-12.)Key Words: systole Ⅲ diastole Ⅲ body mass index Ⅲ genetics Ⅲ risk factors Ⅲ twins H igh blood pressure and body mass index (BMI) predispose independently to death from coronary artery disease and stroke. 1,2 However, hypertension and obesity also often coincide, and cardiovascular risk is augmented under these circumstances. The coincidence of these clinical conditions reflects the underlying correlation between blood pressure and body weight that has been observed at different ages and in a variety of populations. [3][4][5][6] Why there is an association between blood pressure and weight is unclear, although a number of specific physiological hypotheses have been advanced, 7 including the concept of insulin resistance and the metabolic syndrome. 8,9 If common causes of hypertension and obesity can be identified, they offer potentially important targets for more efficient and effective strategies for reduction in cardiovascular risk.Family and twin studies can provide evidence regarding genetic and environmental influences, not only on variation of individual traits, but also on covariation between traits. Such information will help direct molecular, clinical, and epidemiological searches for specific underlying causes. For example, evidence of genetic factors that determine both blood pressure and BMI might lead to more efficient and targeted molecular searches for the specific gene...