Childhood blood pressure tracking correlations corrected for within-person variability.

The correlation coefficient between initial and subsequent blood pressure (BP) measurements is referred to as the tracking correlation. Childhood BP tracking correlations, although positive, have been considered too low to make accurate predictions for an individual. These correlations, however, can be raised substantially by averaging BP over multiple weekly visits in each year, which partially accounts for within-person variability. In a cohort of 333 school children, we measured BP 3 times on each of 4 successive weekly visits, in each of 4 consecutive years, using a random-zero sphygmomanometer. Approximately 90 per cent of subjects had data for one or more follow-up years, and 75 per cent of subjects who entered in the first year had data for all four years. With a model that allows estimation of correlations and that uses all available longitudinal data, we calculated tracking correlations completely corrected for within-person variability, the statistical equivalent of measuring BP on an infinite number (infinity) of visits and measurements per visit. Age-sex adjusted tracking correlations for 3 years of follow-up based on the means from 1,2,3,4, and infinity visits are, for systolic BP, 0.43, 0.56, 0.62, 0.66, and 0.73, respectively, and for diastolic BP, 0.20, 0.37, 0.46, 0.50, and 0.70, respectively. With longer follow-up, the use of corrected tracking correlations would allow determination of the maximal extent to which childhood BP can predict adult levels, and therefore the usefulness of screening to identify children at high risk of developing hypertension.     

Combining annual blood pressure measurements in childhood to improve prediction of young adult blood pressure

Tracking correlations of blood pressure (BP) have been reported between levels measured in a single year during both childhood and adulthood. Because of the variability of BP, these correlations increase with the number of visits and measurements per visit in each year. It remains unclear, however, whether such correlations would improve further by combining BP data collected over several years. From 1978-1981, BP was measured annually in a cohort of 339 children in East Boston, MA, at four visits one week apart with three measurements per visit. Of this cohort, then aged 18-26 years, 316 were re-examined in 1989-1990 at three visits one week apart with three measures per visit. Tracking correlations were estimated from levels measured in a single year as well as means averaged over several years in childhood, adjusting for age, year of measurement, as well as smoking, alcohol and oral contraceptive use. Multivariate models were fit to estimate tracking correlations from childhood to young adulthood adjusting for within-person variability. Using a single year in childhood, these were 0.49 in boys and 0.59 in girls for systolic BP and 0.39 and 0.48 for diastolic BP (all p<0.001). Using the long-term average in childhood and adjusting for variability across years, these values were 0.55 in boys and 0.66 in girls for systolic BP and 0.47 and 0.57 for diastolic BP (all p<0.001). We observed concomitant increases in the predictive value of childhood BP for young adult BP. These results suggest that averaging BP over at least two years during childhood increases tracking correlations and improves the prediction of adult values from childhood levels.     

Serial changes in blood pressure from adolescence into adulthood.

High blood pressure is an established risk factor for cardiovascular disease outcomes in adulthood. Furthermore, numerous longitudinal studies of blood pressure in childhood with length of follow-up from 1 to 17 years indicate that blood pressure levels track over the short term. This study addresses the question of the predictive value of childhood blood pressure readings for adult levels, using repeated blood pressure determinations from a sample of 501 participants in the Fels Longitudinal Study, an ongoing cohort study in southwestern Ohio that began in 1929. A damped autoregressive model indicated tracking correlations from 0.39 (4-year intervals) to 0.24 (20 years) for systolic pressure and 0.37 (4 years) to 0.20 (20 years) for diastolic pressure. These results indicate that tracking of blood pressure persists from age 13 years to age 40 years, which translates into moderate levels of relative risk for adult hypertension (diastolic pressure above 90 mmHg) for adolescents with high normal blood pressure. The estimated relative risks of hypertension at age 35 for white 15-year-olds with a true mean diastolic pressure of 80 mmHg were 1.9 for males and 2.6 for females, relative to 15-year-olds with a true diastolic pressure of 60 mmHg.     

ESTIMATING PREDICTIVE VALUES FOR BLOOD PRESSURE MEASUREMENTS FROM MULTIVARIATE REGRESSION MODELS WITH COVARIATES

Predictive values are useful in estimating the probability distribution of a ‘true’ or underlying measurement, that is, without measurement error or within-person variability. They have been applied to blood pressure data to estimate the true probability that a person is hypertensive currently, or that he/she will become hypertensive based on previous data from childhood. The current work extends these results to situations where covariates are of interest. One can use multivariate regression models to model predictive values for future levels as functions of covariates as well as current measured levels. I compare predictive value estimates obtained from these models to those obtained from ordinary linear regression and from logistic regression with use of data on childhood blood pressure from East Boston, MA. Estimates obtained using the multivariate model are preferable either in terms of bias in the estimates themselves or in terms of their variability. This is particularly true with covariates included in the model. The difference between the multivariate and ordinary regression estimates depends on the conditional reliability of future levels given current blood pressure levels and covariates. I also discuss predictive value estimates for true current level given observed level as well as covariates. These also depend on the reliability of the current measure given values of covariates.     

Using the area under the curve to reduce measurement error in predicting young adult blood pressure from childhood measures

Tracking correlations of blood pressure, particularly childhood measures, may be attenuated by within-person variability. Combining multiple measurements can reduce this error substantially. The area under the curve (AUC) computed from longitudinal growth curve models can be used to improve the prediction of young adult blood pressure from childhood measures. Quadratic random-effects models over unequally spaced repeated measures were used to compute the area under the curve separately within the age periods 5-14 and 20-34 years in the Bogalusa Heart Study. This method adjusts for the uneven age distribution and captures the underlying or average blood pressure, leading to improved estimates of correlation and risk prediction. Tracking correlations were computed by race and gender, and were approximately 0.6 for systolic, 0.5-0.6 for K4 diastolic, and 0.4-0.6 for K5 diastolic blood pressure. The AUC can also be used to regress young adult blood pressure on childhood blood pressure and childhood and young adult body mass index (BMI). In these data, while childhood blood pressure and young adult BMI were generally directly predictive of young adult blood pressure, childhood BMI was negatively correlated with young adult blood pressure when childhood blood pressure was in the model. In addition, racial differences in young adult blood pressure were reduced, but not eliminated, after controlling for childhood blood pressure, childhood BMI, and young adult BMI, suggesting that other genetic or lifestyle factors contribute to this difference.     

Reproducibility and predictive values of routine blood pressure measurements in children. Comparison with adult values and implications for screening children for elevated blood pressure

A study of the variability of blood pressure was conducted among a total of 780 Massachusetts children, 335 children in East Boston and 445 children in Brookline, ages 8-18 years. All children had their blood pressure measured with a standard mercury sphygmomanometer in a school setting on four visits one week apart with three measurements per visit. In East Boston, repeat measurements were made for the same children for four consecutive years. A nested random effects model was used to estimate between- and within-visit variance components. For children aged 8-12 years, these were, respectively, 33.1, 12.0 in boys and 31.2, 11.1 in girls for systolic blood pressure and 57.7, 21.3 in boys and 56.6, 22.6 in girls for systolic muffling blood pressure (Korotkoff phase 4). For children aged 13-18 years of age, they were, respectively, 41.1, 11.8 in boys and 35.2, 12.2 in girls for blood pressure and 40.6, 15.5 in boys and 36.1, 11.4 in girls for diastolic blood pressure (Korotkoff phase 5). Within-person variability for systolic pressure was comparable to previously published data for 434 white adults ages 30-49 years not on antihypertensive medications; however, within-person variability for diastolic pressure was considerably higher in the children, accounting for over 75% of total variability among 8-12-year-old children, compared with 27% for adults. No meaningful effects of age, sex, or blood pressure level on variability of systolic pressure were found. However, age and level of blood pressure each had a large and independent inverse association with variability of diastolic pressure; variance components for younger children (ages 8-12 years) and children with low diastolic pressure (less than 60 mmHg) were approximately twice as large as for older children (ages 13-18 years) and children with diastolic pressure greater than or equal to 60 mmHg, respectively. Finally, predictive value estimates of blood pressure are provided for particular age-sex groups to enable one to efficiently identify children whose true mean level of blood pressure exceeds the 90th percentile for their age-sex group with minimum misclassification. Because of the substantial variability of diastolic pressure in young children, resulting in relatively low predictive value estimates, systolic pressure (either alone or in combination with diastolic pressure) may be more useful as the primary tool for screening children under age 13 years for high blood pressure.     

Meta‐analysis of diagnostic accuracy studies accounting for disease prevalence: Alternative parameterizations and model selection

In a meta‐analysis of diagnostic accuracy studies, the sensitivities and specificities of a diagnostic test may depend on the disease prevalence since the severity and definition of disease may differ from study to study due to the design and the population considered. In this paper, we extend the bivariate nonlinear random effects model on sensitivities and specificities to jointly model the disease prevalence, sensitivities and specificities using trivariate nonlinear random‐effects models. Furthermore, as an alternative parameterization, we also propose jointly modeling the test prevalence and the predictive values, which reflect the clinical utility of a diagnostic test. These models allow investigators to study the complex relationship among the disease prevalence, sensitivities and specificities; or among test prevalence and the predictive values, which can reveal hidden information about test performance. We illustrate the proposed two approaches by reanalyzing the data from a meta‐analysis of radiological evaluation of lymph node metastases in patients with cervical cancer and a simulation study. The latter illustrates the importance of carefully choosing an appropriate normality assumption for the disease prevalence, sensitivities and specificities, or the test prevalence and the predictive values. In practice, it is recommended to use model selection techniques to identify a best‐fitting model for making statistical inference. In summary, the proposed trivariate random effects models are novel and can be very useful in practice for meta‐analysis of diagnostic accuracy studies. Copyright © 2009 John Wiley & Sons, Ltd.     

Comparing new and old screening tests when a reference procedure cannot be performed on all screenees. Example of automated cytometry for early detection of cervical cancer

Direct determination of the sensitivity and specificity of a screening test requires use of a reference procedure (such as biopsy with histopathologic analysis) that provides an estimate of true disease status. The authors present a method for comparing the accuracy of a new screening test to an old one in situations when it is not feasible to apply the reference procedure to all screenees. This method requires that only those persons who test positive on old or new screening tests be further evaluated with the reference procedure. Ratios of sensitivities and specificities are derived for rapid comparison of the two screening tests. It is shown that McNemar's test can be used for significance testing of the differences in sensitivities and specificities between two screening tests. The required sample size for a study that compares the two tests is determined.     

Optimal use of literature knowledge to improve the Bayesian diagnosis of coronary artery disease

Bayes' theorem with the independence assumption is applied to a test sample of 141 subjects, using two sets of test sensitivities and specificities. The first set is derived by averaging over literature reports on the accuracy of the exercise electrocardiogram, exercise thallium scintigraphy, and carciac fluoroscopy. The second set of indices is derived by applying multivariate regression to the technical, population, and methodologic attributes obtained from the same literature by the use of meta-analysis. The meta-analytically corrected sensitivities and specificities resulted in significant improvement in the discriminatory power of the Bayes model. (Area under ROC curve increased, p = less than 0.01). However, the corrected model was not as accurate as a data-derived logistic regression model of the same test variables. Meta-analysis may be useful for modest improvement in the accuracy of literature-derived Bayesian models for predicting disease probabilities.     

Home monitoring of blood pressure: usefulness as a predictor of persistent hypertension

We set out to test the hypothesis that home blood pressure reflects "baseline" pressures measured at a general practitioner's surgery or in a hospital outpatient clinic. Twenty patients detected hypertensive during screening in general practice and 30 patients referred to a hospital hypertension clinic for revision of therapy were studied. All were instructed in the use of an electronic semiautomatic sphygmomanometer and measured blood pressure at home for a three day period. Home monitored blood pressure correctly predicted those patients whose diastolic blood pressure fell to below 95 mmHg by the third clinic visit in approximately 90% of all patients. In addition, in those whose blood pressure was high at home it remained so at the clinic or surgery after three visits. These data suggest that home monitoring of blood pressure may be a helpful alternative to repeated clinic visits before embarking on medical therapy.     

采用血压身高比筛查广州市部分儿童高血压的初步探索

目的了解广州市儿童的高血压水平,探索采用血压身高比筛查广州市儿童高血压的最佳阈值。方法采用整群随机抽样的方法对广州市6~10岁的5 471名儿童进行血压及相关生长发育指标测量,采用受试者工作特征曲线确定血压身高比预测儿童高血压的能力及最佳阈值。结果用于筛查广州市儿童高血压的收缩压身高比的最佳阈值为男生0.80mmHg/cm,女生0.78mmHg/cm;舒张压身高比的最佳阈值为男生0.51mmHg/cm,女生0.49mmHg/cm,阴性预测值高于阳性预测值,且均大于99%,阳性预测值较低,男生为17.2%,女生为15.9%。入组儿童高血压发病率为3.60%,其中男3.54%,女3.67%。结论血压身高比可作为一种简单精确的指标筛查广州市6~10岁儿童高血压。     

国产HIV抗体检测试剂盒的评价

本文报作了对目前国产的HIV抗体检测粗筛试剂的评价结果。病毒所生产的IF和IE试剂,及上海生研所生产的ELISA试剂,敏感性为91.2%~96.9%;特异性为94.6%~97.3%。在我国这样的HIV低感染率情况下,这些试剂的阴性预测值可达100%,阳性预测值极低。认为这些试剂都还能满足国内HIV检测粗筛实验的应用。这些试剂的剂型价格,尚须改进。     

Comparison of diagnostic markers with repeated measurements: a non-parametric ROC curve approach

In this paper we study a class of non-parametric statistics for comparing diagnostic markers with repeated measurements. Using adapted definitions of specificity and sensitivity, we suggest methods to compare the average of sensitivities across all specificities or a range of specificities. The theory allows for correlations introduced by the fact that markers may be obtained from the same patient at multiple visits and that both markers being compared may be obtained from the same patient. Results of the Monte Carlo simulations and an example from a breast cancer setting are provided.     

三种宫颈癌联合筛查方案的对比研究

目的:研究三种宫颈癌联合筛查方案的临床价值。方法:选择江西省人民医院妇科健康体检3 280例妇女,随机分为三组:A组(1 100例)行TCT联合HPV-DNA分型检测(PCR-反向点杂交法);B组(1 080例)行巴氏涂片法联合HPV-DNA检测(PCR荧光定性);C组(1 100例)行醋酸染色(VIA)、碘液染色(VILI)肉眼观察。比较三种联合方案对宫颈病变筛查的灵敏度、特异度、阳性预测值、阴性预测值。结果:三组筛查方案的灵敏度、特异度、阳性预测值、阴性预测值分别为:A组:78.9%、98.3%、42.9%、99.6%;B组:25.0%、98%、33.3%、98.5%;C组:73.7%、91.1%、12.7%、99.5%。三组比较差异有统计学意义(P<0.05)。结论:TCT联合HPV-DNA分型检测筛查方案诊断价值优于其余两种方案,可作为宫颈癌筛查的最佳方案,适用于经济发达地区;巴氏涂片法联合HPV-DNA检测诊断准确性高,适用于一般地区;肉眼观察,简单易行,价格低廉,可在经济欠发达地区农村基层推广应用。     

A study of the correlation between serum total cholesterol and low-density lipoproteins (LDL) in Chinese

Serum total cholesterol and lipoprotein analysis was performed on the same blood samples of 139 asymptomatic Chinese subjects aged 40-60 years. There was a highly significant correlation between serum total cholesterol and low density lipoproteins (correlation coefficient = 0.8376, p less than 0.001) in Chinese, which means that total cholesterol can be used as the initial screening to identify individuals who may need lipoprotein analysis. The American National Cholesterol Education Programme (ANCEP) recommended threshold total cholesterol levels that indicate lipoprotein analysis for the high risk and low risk groups were found to have high sensitivities and negative predictive values but relatively low specificities and positive predictive values. Thirty to forty per cent of the lipoprotein analyses recommended by the ANCEP guidelines were not necessary. There was no significant age or sex influence on the degree of correlation between serum total cholesterol and LDL.     

Optimal two‐phase sampling design for comparing accuracies of two binary classification rules

In this paper, we consider the design for comparing the performance of two binary classification rules, for example, two record linkage algorithms or two screening tests. Statistical methods are well developed for comparing these accuracy measures when the gold standard is available for every unit in the sample, or in a two‐phase study when the gold standard is ascertained only in the second phase in a subsample using a fixed sampling scheme. However, these methods do not attempt to optimize the sampling scheme to minimize the variance of the estimators of interest. In comparing the performance of two classification rules, the parameters of primary interest are the difference in sensitivities, specificities, and positive predictive values. We derived the analytic variance formulas for these parameter estimates and used them to obtain the optimal sampling design. The efficiency of the optimal sampling design is evaluated through an empirical investigation that compares the optimal sampling with simple random sampling and with proportional allocation. Results of the empirical study show that the optimal sampling design is similar for estimating the difference in sensitivities and in specificities, and both achieve a substantial amount of variance reduction with an over‐sample of subjects with discordant results and under‐sample of subjects with concordant results. A heuristic rule is recommended when there is no prior knowledge of individual sensitivities and specificities, or the prevalence of the true positive findings in the study population. The optimal sampling is applied to a real‐world example in record linkage to evaluate the difference in classification accuracy of two matching algorithms. Copyright © 2013 John Wiley & Sons, Ltd.     

Algorithms for converting random-zero to automated oscillometric blood pressure values, and vice versa

Many surveys and cohort studies have used a random-zero sphygmomanometer blood pressure device (RZS) to measure subjects' blood pressure and to assess the value of blood pressure in predicting cardiovascular events. Recent studies used automated oscillometric blood pressure devices (AODs) that systematically measure higher blood pressure values than RZSs do, hampering comparability of values between these studies. In 2000-2003, the authors randomly used both an RZS and an AOD in an ongoing cohort study in Germany. This analysis aimed to compare blood pressure values by device and to develop an algorithm to convert estimates of blood pressure values from one device to the other. In a randomized subset of 2,365 subjects aged 45-75 years, each subject was measured three times with each device in a randomized order. The mean difference (AOD-RZS) between the devices was 3.9 mmHg for systolic blood pressure and 2.6 mmHg for diastolic blood pressure. The authors found that linear regression models including age, sex, and blood pressure level can be used to convert RZS blood pressure values to AOD blood pressure values, and vice versa. Results may help to better compare blood pressure values in epidemiologic studies that used different blood pressure devices.     

Test of a noninvasive instrument for measuring hemoglobin concentration

A colorimetric instrument for the noninvasive quantification of hemoglobin, designed using color shades resembling those observed in the conjunctiva, was tested. The instrument's colors are contrasted against the color of the conjunctiva to measure hemoglobin content. Sensitivity, specificity, negative predictive value, and false-negative value were estimated to test the instrument's accuracy; kappa coefficients were used to estimate inter- and intraobserver variability. Physician field evaluations of conjunctiva color for the screening of anemia, reported in the literature, have had sensitivities and specificities as high as 70%. Readings with the instrument demonstrated a 63% sensitivity, 72% specificity, and 38% false negatives for screening hemoglobin values of less than or equal to 13 g/dl. The interobserver kappa coefficients for three pairs of readers were good to excellent for the same hemoglobin screening value. Statistically significant differences were noted, however, between observers during the reliability test. The instrument can be used by unskilled personnel to improve their decision-making about whom to send for further care or for supplementation with iron.     

A five-year follow-up study on blood pressure and serum cholesterol in junior high school children]

Blood pressure and serum cholesterol changes over a five-year period were studied in 299 junior high school children (127 males and 172 females) examined during 1980-1984 in Akita Prefecture, Japan. The six factors studied were height, weight, body mass index (by Minowa's method), systolic blood pressure, diastolic blood pressure and serum cholesterol. Mean values for systolic and diastolic blood pressure increased more than 10 mmHg during the five years in both sexes. Significant positive correlations between initial and follow-up blood pressure ('tracking') were observed. The correlation coefficients for systolic blood pressure in males and in females were 0.33 and 0.28 respectively, and those for diastolic blood pressure were 0.36 in males and 0.19 in females. While there were no significant differences in serum cholesterol levels between the two periods in either sex, the correlation coefficients, which were higher than those for blood pressure, were 0.55 in males and 0.45 in females. Among the six factors at each period, significant positive correlations were observed between height and systolic blood pressure at the initial period, and between obesity and systolic blood pressure at both periods in males and females. A significant positive relationship between obesity and serum cholesterol was seen at the follow-up period in both sexes. These data suggest that a moderate degree of 'tracking' occurs in blood pressure and serum cholesterol during childhood, and that obesity is an important factor related to blood pressure and serum cholesterol.     

Epidemiologic assessment of screening tests for antibody to human T lymphotropic virus type I (HTLV-I).

We tested 196 sera from a human T lymphotropic virus type I (HTLV-I) risk group (prostitute women) with two commercial "research" enzyme-linked immunoabsorbent assays (EIA) for HTLV-I antibodies. All tested sera were characterized by HTLV-I Western immunoblots and by HTLV-I radioimmunoprecipitation assays. The estimated sensitivities of the EIA tests were 93.8 percent and 100 percent, and the specificities were 98.8 percent and 95.8 percent, respectively, using recommended criteria for seropositivity (requiring reactivity to both gag p24 and env gp46 or gp61/68). Calculated negative predictive values remained excellent (greater than 99.9 percent and 100 percent, respectively) at lower seroprevalence rates but the positive predictive values were only 7.3 percent and 2.3 percent when calculated for a seroprevalence rate of 0.1 percent. These results emphasize the importance and need for additional HTLV-I supplementary serologic testing when screening populations with low HTLV-I seroprevalence rates.