Family-based tests of association in the presence of linkage

Linkage analysis may not provide the necessary resolution for identification of the genes underlying phenotypic variation. This is especially true for gene-mapping studies that focus on complex diseases that do not exhibit Mendelian inheritance patterns. One positional genomic strategy involves application of association methodology to areas of identified linkage. Detection of association in the presence of linkage localizes the gene(s) of interest to more-refined regions in the genome than is possible through linkage analysis alone. This strategy introduces a statistical complexity when family-based association tests are used: the marker genotypes among siblings are correlated in linked regions. Ignoring this correlation will compromise the size of the statistical hypothesis test, thus clouding the interpretation of test results. We present a method for computing the expectation of a wide range of association test statistics under the null hypothesis that there is linkage but no association. To standardize the test statistic, an empirical variance-covariance estimator that is robust to the sibling marker-genotype correlation is used. This method is widely applicable: any type of phenotypic measure or family configuration can be used. For example, we analyze a deletion in the A2M gene at the 5' splice site of "exon II" of the bait region in Alzheimer disease (AD) discordant sibships. Since the A2M gene lies in a chromosomal region (chromosome 12p) that consistently has been linked to AD, association tests should be conducted under the null hypothesis that there is linkage but no association.

[Validity and reproducibility of a questionnaire on physical activity and non-activity for school children in Mexico City]

OBJECTIVE

To assess the validity and reproducibility of a self-reported questionnaire on physical activity and inactivity, developed for children aged 10-14 in Mexico City.

MATERIAL AND METHODS

Between May and December 1996, a self-reported physical activity and inactivity questionnaire was developed and applied twice to a sample of 114 students aged 10 to 14, from a low and middle income population of Mexico City. The children's mothers completed the same questionnaire, and two 24-hour recalls of physical activity were used for comparison. Statistical analysis consisted of central tendency and dispersion measures and Pearson's correlation coefficient.

RESULTS

Correlations between hours per day spent in physical activity and inactivity from the children's questionnaire and the 24-hour recall data, were 0.03 for moderate activity, 0.15 for vigorous activity, and 0.51 (p = 0.001) for watching television, adjusted by age, gender, town, and illness prior to the administration of the questionnaire. Compared to the 24-hour recall data, the questionnaire overestimated the time spent watching television, reading or participating in vigorous activity, and underestimated the time engaged in moderate activity. Statistically significant (p < 0.05) six-month reproducibility values were observed for watching television (r = 0.53), sleeping (r = 0.40), moderate (r = 0.38), and vigorous activity (r = 0.55).

CONCLUSIONS

Among children of Mexico City aged 10-14, the questionnaire showed acceptable validity in estimating the time watching television, and acceptable reproducibility of the time watching television, vigorous and moderate activity.

A unified approach to adjusting association tests for population admixture with arbitrary pedigree structure and arbitrary missing marker information

A general approach to family-based examinations of association between marker alleles and traits is proposed. The approach is based on computing p values by comparing test statistics for association to their conditional distributions given the minimal sufficient statistic under the null hypothesis for the genetic model, sampling plan and population admixture. The approach can be applied with any test statistic, so any kind of phenotype and multi-allelic markers may be examined, and covariates may be included in analyses. By virtue of the conditioning, the approach results in correct type I error probabilities regardless of population admixture, the true genetic model and the sampling strategy. An algorithm for computing the conditional distributions is described, and the results of the algorithm for configurations of nuclear families are presented. The algorithm is applicable with all pedigree structures and all patterns of missing marker allele information.

Probabilistic diagnosis in linkage analysis of bipolar disorder: putting weights on the fringe

We explored the utility of probabilistic weighting of fringe phenotypes in linkage analysis of bipolar disorder for the GAW10 chromosome 18 data. Four liability classes were assigned probabilistic weights based on the estimated probability that the case was a true bipolar. The weights were incorporated in parametric and nonparametric, single and multipoint analyses.

A mixture model for longitudinal data with application to assessment of noncompliance

In clinical trials of a self-administered drug, repeated measures of a laboratory marker, which is affected by study medication and collected in all treatment arms, can provide valuable information on population and individual summaries of compliance. In this paper, we introduce a general finite mixture of nonlinear hierarchical models that allows estimates of component membership probabilities and random effect distributions for longitudinal data arising from multiple subpopulations, such as from noncomplying and complying subgroups in clinical trials. We outline a sampling strategy for fitting these models, which consists of a sequence of Gibbs, Metropolis-Hastings, and reversible jump steps, where the latter is required for switching between component models of different dimensions. Our model is applied to identify noncomplying subjects in the placebo arm of a clinical trial assessing the effectiveness of zidovudine (AZT) in the treatment of patients with HIV, where noncompliance was defined as initiation of AZT during the trial without the investigators' knowledge. We fit a hierarchical nonlinear change-point model for increases in the marker MCV (mean corpuscular volume of erythrocytes) for subjects who noncomply and a constant mean random effects model for those who comply. As part of our fully Bayesian analysis, we assess the sensitivity of conclusions to prior and modeling assumptions and demonstrate how external information and covariates can be incorporated to distinguish subgroups.

Generalized linear mixture models for handling nonignorable dropouts in longitudinal studies

This paper presents a method for analysing longitudinal data when there are dropouts. In particular, we develop a simple method based on generalized linear mixture models for handling nonignorable dropouts for a variety of discrete and continuous outcomes. Statistical inference for the model parameters is based on a generalized estimating equations (GEE) approach (Liang and Zeger, 1986). The proposed method yields estimates of the model parameters that are valid when nonresponse is nonignorable under a variety of assumptions concerning the dropout process. Furthermore, the proposed method can be implemented using widely available statistical software. Finally, an example using data from a clinical trial of contracepting women is used to illustrate the methodology.

Incidence of depression in the Stirling County Study: historical and comparative perspectives

BACKGROUND

The Stirling County Study provides a 40-year perspective on the epidemiology of psychiatric disorders in an adult population in Atlantic Canada. Across samples selected in 1952, 1970 and 1992 current prevalence of depression was stable. This paper concerns time trends in annual incidence as assessed through cohorts selected from the first two samples.

METHODS

Consistent interview data were analysed by a computerized diagnostic algorithm. The cohorts consisted of subjects at risk for a first depression: Cohort-1 (N = 575) was followed 1952-1970; Cohort-2 (N = 639) was followed 1970-1992. Life-table methods were used to calculate incidence rates and proportional hazards procedures were used for statistical assessment.

RESULTS

Average annual incidence of depression was 4.5 per 1000 for Cohort-1 and 3.7 for Cohort-2. Differences by gender, age and time were not statistically significant. The stability of incidence and the similarity of distribution by gender and age in these two cohorts corresponds to findings about the two early samples. In contrast, current prevalence in the recent sample was distributed differently and showed an increase among women under 45 years.

CONCLUSIONS

The stability of the incidence of depression emphasizes the distinctive characteristics of current prevalence in the recent sample and suggests that the dominance of women in rates of depression may have occurred among those born after the Second World War. The results offer partial support for the interpretation of an increase in depression based on retrospective data in other recent studies but they indicate that the increase is specific to women.

A 40-year perspective on the prevalence of depression: the Stirling County Study

BACKGROUND

According to epidemiologic studies that use recall of lifetime episodes, the prevalence of depression is increasing. This report from the Stirling County Study compares rates of current depression among representative samples of adults from a population in Atlantic Canada.

METHODS

Sample sizes were 1003, 1201, and 1396 in 1952, 1970, and 1992, respectively. The depression component of the study's method, the DPAX (DP for depression and AX for anxiety), was employed. The original procedure (DPAX-1) was applied in all years. A revision (DPAX-2) was used in 1970 and 1992. The Diagnostic Interview Schedule (DIS) was also used in 1992.

RESULTS

With the DPAX-1, the overall prevalence of current depression was steady at 5% over the 2 early samples but declined in 1992 because of vernacular changes referring to dysphoria. The DPAX-2 gave a stable overall prevalence of 5% in the 2 recent samples, but indicated that women and younger people were at greater risk in 1992 than in 1970. The DIS, like the DPAX-2, found a current 1992 rate of 5% for major depressive episodes combined with dysthymia. Recalled lifetime rates using the DIS showed the same profile interpreted in other studies as suggesting an increase in depression over time.

CONCLUSIONS

Three samples over a 40-year period showed a stable current prevalence of depression using the DPAX methods that was comparable in 1992 with the current rates using the DIS. This casts doubt on the interpretation that depression is generally increasing. Within the overall steady rate observed in this study, historical change was a matter of redistribution by sex and age, with a higher rate among younger women being of recent origin.

A comparison of diagnostic interviews for depression in the Stirling County study: challenges for psychiatric epidemiology

BACKGROUND

High prevalence rates in psychiatric epidemiologic studies raise questions about whether data-gathering procedures identify transient responses rather than clinical disorders. This issue is explored relevant to depression using data from the Stirling County Study.

METHODS

The study's customary method, the DPAX (DP for depression and AX for anxiety) was compared with the Diagnostic Interview Schedule (DIS), both of which were administered to a sample of 1396 subjects selected in 1992. Reasons for discordance were analyzed, and demographic correlates of responses to questions about dysphoria were examined. These lay-administered interviews were then compared with clinician-administered interviews that used the Structured Clinical Interview for DSM-III-R (SCID) with 139 subjects. The kappa statistic and logistic regression were used for statistical assessment.

RESULTS

For the level of agreement between the DPAX and the DIS for current and lifetime depression, kappa = 0.40 and kappa = 0.33, respectively. Subjects diagnosed only by the DPAX tended to have less education than those diagnosed only by the DIS. Some idioms for dysphoria seemed to work better than others. Using SCID interviews as a clinical standard, the DPAX had 15% sensitivity and 96% specificity and the DIS had 25% sensitivity and 98% specificity.

CONCLUSIONS

Comprehension of an interview can be improved by using multiple questions for dysphoria and a simpler mode of inquiry. Clinician-administered interviews tend to corroborate disorders identified in lay-administered interviews but suggest that survey methods underestimate prevalence. Further research is needed to evaluate the validity of both types of interviews, but evidence from a 16-year follow-up evaluation indicates that depression diagnosed by the DPAX is a serious disorder in terms of morbidity and mortality.

Family-based tests of association and linkage that use unaffected sibs, covariates, and interactions

We extend the methodology for family-based tests of association and linkage to allow for both variation in the phenotypes of subjects and incorporation of covariates into general-score tests of association. We use standard association models for a phenotype and any number of predictors. We then construct a score statistic, using likelihoods for the distribution of phenotype, given genotype. The distribution of the score is computed as a function of offspring genotypes, conditional on parental genotypes and trait values for offspring and parents. This approach provides a natural extension of the transmission/disequilibrium test to any phenotype and to multiple genes or environmental factors and allows the study of gene-gene and gene-environment interaction. When the trait varies among subjects or when covariates are included in the association model, the score statistic depends on one or more nuisance parameters. We suggest two approaches for obtaining parameter estimates: (1) choosing the estimate that minimizes the variance of the test statistic and (2) maximizing the statistic over a nuisance parameter and using a corrected P value. We apply our methods to a sample of families with attention-deficit/hyperactivity disorder and provide examples of how covariates and gene-environment and gene-gene interactions can be incorporated.

Patient risk factors for adverse drug events in hospitalized patients. ADE Prevention Study Group

BACKGROUND

Adverse drug events (ADEs) are common in hospitalized patients, but few empirical data are available regarding the strength of patient risk factors for ADEs.

METHODS

We performed a nested case-control study within a cohort that included 4108 admissions to a stratified random sample of 11 medical and surgical units in 2 tertiary care hospitals during a 6-month period. Analyses were conducted on 2 levels: (1) using a limited set of variables available for all patients using computerized data available from 1 hospital and (2) using a larger set of variables for the case patients and matched controls from both hospitals. Case patients were patients with an ADE, and the matched control for each case patient was the patient on the same unit as the case patient with the most similar prevent length of stay. Main outcome measures were presence of an ADE, preventable ADE, or severe ADE.

RESULTS

In the cohort analysis, electrolyte concentrates (odds ratio [OR], 1.7), diuretics (OR, 1.7), and medical admission (OR, 1.6) were independent correlates of ADEs. Independent correlates of preventable ADEs in the cohort analysis were low platelet count (OR, 4.5), antidepressants (OR, 3.3), antihypertensive agents (OR, 2.9), medical admission (OR, 2.2), and electrolyte concentrates (OR, 2.1). In the case-control analysis, exposure to psychoactive drugs (OR, 2.1) was an independent correlate of an ADE, and use of cardiovascular drugs (OR, 2.4) was independently correlated with severe ADEs. For preventable ADEs, no independent predictors were retained after multivariate analysis.

CONCLUSIONS

Adverse drug events occurred more frequently in sicker patients who stayed in the hospital longer. However, after controlling for level of care and preevent length of stay, few risk factors emerged. These results suggest that, rather than targeting ADE-prone individuals, prevention strategies should focus on improving medication systems.

Correcting for non-compliance in randomized trials: an application to the ATBC Study

Different methods for estimating the effect of treatment actually received in a longitudinal placebo-controlled trial with non-compliance are discussed. Total mortality from the ATBC Study is used as an illustrative example. In the ATBC Study some 25 per cent of the participants dropped out from active follow-up prior to the scheduled end of the study. The 'intention-to-treat' analysis showed an increased death risk in the beta-carotene arm when compared with the no beta-carotene arm. Owing to considerable non-compliance it is also of interest to estimate the effect of beta-carotene actually received. We use a simple model for the treatment action and discuss three methods for estimation of the treatment effect under the model - the 'intention-to-treat' approach, the 'as-treated' approach and the g-estimation approach. These approaches are compared in a simulation study under different settings for non-compliance. Finally, the data from the ATBC Study are analysed using the proposed methods.

Weight cycling, weight gain, and risk of hypertension in women

To assess prospectively the relation between body mass index, weight gain, repeated intentional weight losses, and the risk of self-reported hypertension, the authors studied 46,224 women who were participants in the Nurses Health Study II, who were free of hypertension in 1993, and who completed questions on intentional weight losses between 1989 and 1993. Women who reported they had intentionally lost > or =20 lbs (9 kg) > or =3 times were classified as severe weight cyclers. Women who had intentionally lost > or =10 lbs (4.5 kg) > or =3 times, but who did not meet the criteria for severe weight cycling, were classified as mild weight cyclers. Between 1993 and 1995, 1,107 incident cases of diagnosed hypertension were reported. Body mass index and weight gain, but not weight cycler status, were independently associated with the development of hypertension. For each 10 lb (4.5 kg) gain in weight between 1989 and 1993, the risk of hypertension increased 20% (odds ratio (OR) = 1.20, 95% confidence interval (CI) 1.15, 1.24). After adjustment for body mass index and weight gain, the risks associated with mild weight cycling (OR = 1.15, 95% CI 1.00, 1.33) and severe weight cycling (OR = 1.13, 95% CI 0.79, 1.61) were small and not significant. Thus, the results of this study offer support for the current weight guidelines and provide further evidence of the health risks associated with excessive weight and weight gain. However, these data do not suggest an independent effect of weight cycling on risk of hypertension.

Impact of a school-based interdisciplinary intervention on diet and physical activity among urban primary school children: eat well and keep moving

OBJECTIVE

To evaluate the impact of a school-based interdisciplinary health behavior intervention on diet and physical activity among children in grades 4 and 5.

DESIGN

A quasiexperimental field trial with 6 intervention and 8 matched control schools. Outcomes were assessed longitudinally using preintervention (fall 1995) and follow-up (spring 1997) student survey food frequency and activity measures and follow-up 24-hour recall measures of diet and activity. Change was also assessed using yearly repeated cross-sectional surveys of all grade 5 students from 1995 through 1997.

PARTICIPANTS

Longitudinal data were collected from 479 students initially in grade 4 in Baltimore, Md, public schools; 91% were African American. Repeated 24-hour recall measures in 1997 were collected for a random subsample of 336 students. Cross-sectional survey data were collected from all grade 5 students in 1995,1996, and 1997 (n = 2103).

INTERVENTION

The Eat Well and Keep Moving Program was taught by classroom teachers over 2 years in math, science, language arts, and social studies classes. Materials provided links to school food services and families and provided training and wellness programs for teachers and other staff members. Intervention materials focused on decreasing consumption of foods high in total and saturated fat and increasing fruit and vegetable intake, as well as reducing television viewing and increasing physical activity.

MAIN OUTCOME MEASURES

Dietary intake and physical activity measured via repeated 24-hour recall were primary end points, with additional food frequency and activity measures.

RESULTS

The 24-hour recall measures indicated that, after controlling for baseline covariates, the percentages of total energy from fat and saturated fat were reduced among students in intervention compared with control schools (-1.4%; 95% confidence interval [CI], -2.8 to -0.04; P = .04 and -0.60%; 95% CI, -1.2 to -0.01; P = .05). There was an increase in fruit and vegetable intake (0.36 servings/4184 kJ; 95% CI, 0.10-0.62; P=.01), in vitamin C intake (8.8 mg/4184 kJ; 95% CI, 2.0-16; P=.01), and in fiber consumption (0.7 g/4184 kJ; 95% CI, 0.0-1.4; P=.05). Television viewing was marginally reduced (-0.55 h/d; 95% CI, -1.04 to 0.04; P=.06). Analysis of longitudinal and repeated cross-sectional food frequency data indicated similar significant decreases in the percentages of total energy from fat and saturated fat.

CONCLUSION

Evaluation of the Eat Well and Keep Moving Program indicates effectiveness in improving dietary intake of students and reducing television viewing.

Association of obesity with physical activity, television programs and other forms of video viewing among children in Mexico city

OBJECTIVE

To assess the association of physical activity, television program viewing and other forms of video viewing with the prevalence of obesity among school children.

DESIGN

Cross-sectional study.

SUBJECTS

712 children, 9-16 y old, from a low- and a middle-income town in the Mexico City area.

MEASUREMENTS

Children completed a self administered questionnaire to assess time spent in physical activity and television viewing, and diet. Height weight and triceps skinfolds were measured. The outcome variable was obesity, and the covariates were hours of television programs and other video viewing, physical activity, energy intake, percentage of energy from fat, town of location of school, age, gender and perception of mother's weight status.

RESULTS

Among 461 children with complete information, 24% were classified as obese. Children reported an average of 4.1 +/- 2.2 h/d watching television (2.4 +/- 1.5 h/d for TV programs and 1.7 +/- 1.5 h/d for video cassette recorder (VCR) or videogames), and 1.8 +/- 1.3 h/d in moderate and vigorous physical activities. Odds ratios (OR) of obesity were 12% higher for each hour of television program viewing per day (OR = 1.12, 95% confidence interval (CI) 1.02,-1.22), and 10% lower for each hour of moderate/vigorous physical activity per day (OR = 0.90, 95% CI 0.83-0.98), controlling for age, gender, town and perception of mother's weight status. Children in the middle-income town had higher adjusted odds of obesity (OR = 2.58, 95% CI 1.47-4.54).

CONCLUSION

Physical activity and television viewing, but not VCR/videogames use, were related to obesity prevalence in Mexican children 9-16 y old.

Reducing obesity via a school-based interdisciplinary intervention among youth: Planet Health

OBJECTIVE

To evaluate the impact of a school-based health behavior intervention known as Planet Health on obesity among boys and girls in grades 6 to 8.

DESIGN

Randomized, controlled field trial with 5 intervention and 5 control schools. Outcomes were assessed using preintervention (fall 1995) and follow-up measures (spring 1997), including prevalence, incidence, and remission of obesity.

PARTICIPANTS

A group of 1295 ethnically diverse grade 6 and 7 students from public schools in 4 Massachusetts communities.

INTERVENTION

Students participated in a school-based interdisciplinary intervention over 2 school years. Planet Health sessions were included within existing curricula using classroom teachers in 4 major subjects and physical education. Sessions focused on decreasing television viewing, decreasing consumption of high-fat foods, increasing fruit and vegetable intake, and increasing moderate and vigorous physical activity.

MAIN OUTCOME MEASURES

Obesity was defined as a composite indicator based on both a body mass index and a triceps skinfold value greater than or equal to age- and sex-specific 85th percentiles. Because schools were randomized, rather than students, the generalized estimating equation method was used to adjust for individual-level covariates under cluster randomization.

RESULTS

The prevalence of obesity among girls in intervention schools was reduced compared with controls, controlling for baseline obesity (odds ratio, 0.47; 95% confidence interval, 0.24-0.93; P = .03), with no differences found among boys. There was greater remission of obesity among intervention girls vs. control girls (odds ratio, 2.16; 95% confidence interval, 1.07-4.35; P = .04). The intervention reduced television hours among both girls and boys, and increased fruit and vegetable consumption and resulted in a smaller increment in total energy intake among girls. Reductions in television viewing predicted obesity change and mediated the intervention effect. Among girls, each hour of reduction in television viewing predicted reduced obesity prevalence (odds ratio, 0.85; 95% confidence interval, 0.75-0.97; P = .02).

CONCLUSION

Planet Health decreased obesity among female students, indicating a promising school-based approach to reducing obesity among youth.

Maximum likelihood analysis of generalized linear models with missing covariates

Missing data is a common occurrence in most medical research data collection enterprises. There is an extensive literature concerning missing data, much of which has focused on missing outcomes. Covariates in regression models are often missing, particularly if information is being collected from multiple sources. The method of weights is an implementation of the EM algorithm for general maximum-likelihood analysis of regression models, including generalized linear models (GLMs) with incomplete covariates. In this paper, we will describe the method of weights in detail, illustrate its application with several examples, discuss its advantages and limitations, and review extensions and applications of the method.

Angiotensinogen gene polymorphisms M235T/T174M: no excess transmission to hypertensive Chinese

The gene encoding angiotensinogen (AGT) has been widely studied as a candidate gene for hypertension. Most studies to date have relied on case-control analysis to test for an excess of AGT variants among hypertensive cases compared with normotensive controls. However, with this design, nothing guarantees that a positive finding is due to actual allelic association as opposed to an inappropriate control population. To avoid this difficulty in our study of essential hypertension in Anqing, China, we tested AGT variants using the transmission/disequilibrium test, a procedure that bypasses the need for a control sample by testing for excessive transmission of a genetic variant from parents heterozygous for that variant. We analyzed two AGT polymorphisms, M235T and T174M, which have been associated with essential hypertension in whites and Japanese, using data on 335 hypertensive subjects from 315 nuclear families and their parents. Except in the group of subjects younger than 25 years, M235 and T174 were the more frequently transmitted alleles. We found that 194 parents heterozygous for M235T transmitted M235 106 times (P=0.22) and that 102 parents heterozygous for T174M transmitted T174 60 times (P=0.09). Stratifying offspring by gender, M235 and T174 were transmitted 60 of 106 times (P=0.21) and 44 of 75 times (P=0.17), respectively, in men, and 46 of 88 times (P=0.75) and 16 of 27 times (P=0.44), respectively, in women. Our results were also negative in all age groups and for the affected offspring with blood pressure values >/=160/95 mm Hg. Thus, this study provides no evidence that either allele of M235T or T174M contributes to hypertension in this Chinese population.

A discordant-sibship test for disequilibrium and linkage: no need for parental data

The sibship disequilibrium test (SDT) is designed to detect both linkage in the presence of association and association in the presence of linkage (linkage disequilibrium). The test does not require parental data but requires discordant sibships with at least one affected and one unaffected sibling. The SDT has many desirable properties: it uses all the siblings in the sibship; it remains valid if there are misclassifications of the affectation status; it does not detect spurious associations due to population stratification; asymptotically it has a chi2 distribution under the null hypothesis; and exact P values can be easily computed for a biallelic marker. We show how to extend the SDT to markers with multiple alleles and how to combine families with parents and data from discordant sibships. We discuss the power of the test by presenting sample-size calculations involving a complex disease model, and we present formulas for the asymptotic relative efficiency (which is approximately the ratio of sample sizes) between SDT and the transmission/disequilibrium test (TDT) for special family structures. For sib pairs, we compare the SDT to a test proposed both by Curtis and, independently, by Spielman and Ewens. We show that, for discordant sib pairs, the SDT has good power for testing linkage disequilibrium relative both to Curtis's tests and to the TDT using trios comprising an affected sib and its parents. With additional sibs, we show that the SDT can be more powerful than the TDT for testing linkage disequilibrium, especially for disease prevalence >.3.

Effect of computerized physician order entry and a team intervention on prevention of serious medication errors

CONTEXT

Adverse drug events (ADEs) are a significant and costly cause of injury during hospitalization.

OBJECTIVES

To evaluate the efficacy of 2 interventions for preventing nonintercepted serious medication errors, defined as those that either resulted in or had potential to result in an ADE and were not intercepted before reaching the patient.

DESIGN

Before-after comparison between phase 1 (baseline) and phase 2 (after intervention was implemented) and, within phase 2, a randomized comparison between physician computer order entry (POE) and the combination of POE plus a team intervention.

SETTING

Large tertiary care hospital.

PARTICIPANTS

For the comparison of phase 1 and 2, all patients admitted to a stratified random sample of 6 medical and surgical units in a tertiary care hospital over a 6-month period, and for the randomized comparison during phase 2, all patients admitted to the same units and 2 randomly selected additional units over a subsequent 9-month period.

INTERVENTIONS

A physician computer order entry system (POE) for all units and a team-based intervention that included changing the role of pharmacists, implemented for half the units.

MAIN OUTCOME MEASURE

Nonintercepted serious medication errors.

RESULTS

Comparing identical units between phases 1 and 2, nonintercepted serious medication errors decreased 55%, from 10.7 events per 1000 patient-days to 4.86 events per 1000 (P=.01). The decline occurred for all stages of the medication-use process. Preventable ADEs declined 17% from 4.69 to 3.88 (P=.37), while nonintercepted potential ADEs declined 84% from 5.99 to 0.98 per 1000 patient-days (P=.002). When POE-only was compared with the POE plus team intervention combined, the team intervention conferred no additional benefit over POE.

CONCLUSIONS

Physician computer order entry decreased the rate of nonintercepted serious medication errors by more than half, although this decrease was larger for potential ADEs than for errors that actually resulted in an ADE.

Alpha-2 macroglobulin is genetically associated with Alzheimer disease

Alpha-2-macroglobulin (alpha-2M; encoded by the gene A2M) is a serum pan-protease inhibitor that has been implicated in Alzheimer disease (AD) based on its ability to mediate the clearance and degradation of A beta, the major component of beta-amyloid deposits. Analysis of a deletion in the A2M gene at the 5' splice site of 'exon II' of the bait region (exon 18) revealed that inheritance of the deletion (A2M-2) confers increased risk for AD (Mantel-Haenzel odds ratio=3.56, P=0.001). The sibship disequilibrium test (SDT) also revealed a significant association between A2M and AD (P=0.00009). These values were comparable to those obtained for the APOE-epsilon4 allele in the same sample, but in contrast to APOE-epsilon4, A2M-2 did not affect age of onset. The observed association of A2M with AD did not appear to account for the previously published linkage of AD to chromosome 12, which we were unable to confirm in this sample. A2M, LRP1 (encoding the alpha-2M receptor) and the genes for two other LRP ligands, APOE and APP (encoding the amyloid beta-protein precursor), have now all been genetically linked to AD, suggesting that these proteins may participate in a common neuropathogenic pathway leading to AD.

A Bayesian framework for intent-to-treat analysis with missing data

In longitudinal clinical trials, one analysis of interest is an intention-to-treat analysis, which groups subjects according to the randomized treatment regardless of whether they stayed on that treatment or not. When in addition to going off the randomized treatment subjects may also drop out of the study and be lost to follow-up, it is unclear what an intention-to-treat analysis should be. If measurements are made after treatment drop-out on a random sample of subjects who drop the treatment, then Hogan and Laird (1996, Biometrics 52, 1002-1017) present a random effects model, well suited to this type of analysis, which fits a two-piece linear spline to the data with the knot at the time the assigned treatment is dropped. This article presents a Bayesian approach to fitting a similar two-piece linear spline model and shows how the model can be applied to data that have no off-treatment observations.

Increasing efficiency from censored survival data by using random effects to model longitudinal covariates

When estimating a survival time distribution, the loss of information due to right censoring results in a loss of efficiency in the estimator. In many circumstances, however, repeated measurements on a longitudinal process which is associated with survival time are made throughout the observation time, and these measurements may be used to recover information lost to censoring. For example, patients in an AIDS clinical trial may be measured at regular intervals on CD4 count and viral load. We describe a model for the joint distribution of a survival time and a repeated measures process. The joint distribution is specified by linking the survival time to subject-specific random effects characterizing the repeated measures, and is similar in form to the pattern mixture model for multivariate data with nonignorable nonresponse. We also describe an estimator of survival derived from this model. We apply the methods to a long-term AIDS clinical trial, and study properties of the survival estimator. Monte Carlo simulation is used to estimate gains in efficiency when the survival time is related to the location and scale of the random effects distribution. Under relatively light censoring (20%), the methods yield a modest gain in efficiency for estimating three-year survival in the AIDS clinical trial. Our simulation study, which mimics characteristics of the clinical trial, indicates that much larger gains in efficiency can be realized under heavier censoring or with studies designed for long term follow up on survival.

Tests for homogeneity of the risk difference when data are sparse

Test statistics for the homogeneity of the risk difference for a series of 2 x 2 tables when the data are sparse is proposed. A weighted least squares statistic is commonly used to test for equality of the risk difference over the tables; however, when the data are sparse, this statistic can have anticonservative Type I error rates. Simulation is used to compare the proposed test statistics to the weighted least squares statistic. The weighted least squares statistic has the most anticonservative Type I error rates of all the statistics compared. We suggest the use of one of our proposed test statistics instead of the weighted least squares statistic.