Almost efficient estimation of relative risk regression.

Relative risks (RRs) are often considered the preferred measures of association in prospective studies, especially when the binary outcome of interest is common. In particular, many researchers regard RRs to be more intuitively interpretable than odds ratios. Although RR regression is a special case of generalized linear models, specifically with a log link function for the binomial (or Bernoulli) outcome, the resulting log-binomial regression does not respect the natural parameter constraints. Because log-binomial regression does not ensure that predicted probabilities are mapped to the [0,1] range, maximum likelihood (ML) estimation is often subject to numerical instability that leads to convergence problems. To circumvent these problems, a number of alternative approaches for estimating RR regression parameters have been proposed. One approach that has been widely studied is the use of Poisson regression estimating equations. The estimating equations for Poisson regression yield consistent, albeit inefficient, estimators of the RR regression parameters. We consider the relative efficiency of the Poisson regression estimator and develop an alternative, almost efficient estimator for the RR regression parameters. The proposed method uses near-optimal weights based on a Maclaurin series (Taylor series expanded around zero) approximation to the true Bernoulli or binomial weight function. This yields an almost efficient estimator while avoiding convergence problems. We examine the asymptotic relative efficiency of the proposed estimator for an increase in the number of terms in the series. Using simulations, we demonstrate the potential for convergence problems with standard ML estimation of the log-binomial regression model and illustrate how this is overcome using the proposed estimator. We apply the proposed estimator to a study of predictors of pre-operative use of beta blockers among patients undergoing colorectal surgery after diagnosis of colon cancer.

Using the jackknife for estimation in log link Bernoulli regression models.

Bernoulli (or binomial) regression using a generalized linear model with a log link function, where the exponentiated regression parameters have interpretation as relative risks, is often more appropriate than logistic regression for prospective studies with common outcomes. In particular, many researchers regard relative risks to be more intuitively interpretable than odds ratios. However, for the log link, when the outcome is very prevalent, the likelihood may not have a unique maximum. To circumvent this problem, a 'COPY method' has been proposed, which is equivalent to creating for each subject an additional observation with the same covariates except the response variable has the outcome values interchanged (1's changed to 0's and 0's changed to 1's). The original response is given weight close to 1, while the new observation is given a positive weight close to 0; this approach always leads to convergence of the maximum likelihood algorithm, except for problems with convergence due to multicollinearity among covariates. Even though this method produces a unique maximum, when the outcome is very prevalent, and/or the sample size is relatively small, the COPY method can yield biased estimates. Here, we propose using the jackknife as a bias-reduction approach for the COPY method. The proposed method is motivated by a study of patients undergoing colorectal cancer surgery.

Communication practices on 4 Harvard surgical services: a surgical safety collaborative.

BACKGROUND: Communication breakdowns between surgical residents and attending physicians in the pre- and postoperative setting are common contributors to patient injury. These communication transactions might offer an opportunity for safety improvement, but it remains unknown how often resident-attending communication fails, what the current level of attending involvement is, and how often attending input changes the plan for patient care. We conducted a prospective study at 4 Harvard teaching hospitals to address these issues. METHODS: Three prospective data collection strategies were employed: (1) we randomly selected surgical services and queried residents for the occurrence of predefined critical patient events and the characteristics of attending communications that ensued, (2) on weekends, randomly selected patients were interviewed and their charts reviewed to identify the frequency of attending visitation and how such visits affected processes of care, and (3) on weekends, senior residents on randomly selected surgical services were queried regarding the occurrence of attending-resident discussion of patients in their care. RESULTS: Of 80 critical patient events identified, 26 (33%) were not communicated to attending surgeons. Residents reported that, when contacted, all attending physicians were receptive to communication, whether they were the primary surgeon or providing cross-coverage. Although residents felt that attending contact was unnecessary for safe patient care in 61 (76%) of these events, discussions with attending physicians changed management in 33% (18/54) of cases in which they occurred. Attending surgeons were found to visit their patients on randomly selected weekend days 42% (n = 37) of the time, while 21% (n = 19) had not visited for 2 or greater days. When attending physicians visited patients, however, resident management was modified 46% (n = 36) of the time. Though residents frequently discussed patient management with attending physicians on randomly selected weekends, they failed to do so 16% (n = 58) of the time, which appeared to be related to service-specific variation (chi = 269, P < 0.0001). CONCLUSIONS: In the context of both critical patient events and routine patient care, residents often fail to obtain attending surgeons' input for management decisions. These failures seem to derive more from residents' perception of necessity than from attending physicians' receptiveness or interest in being contacted. Once involved, attending physicians frequently modify resident's management decisions. It seems, therefore, that there is significant potential for communication failure and information loss among our 4 institutions.

Expanded genomewide scan implicates a novel locus at 3q28 among Caribbean hispanics with familial Alzheimer disease.

OBJECTIVES: To identify novel candidate regions for late-onset Alzheimer disease (LOAD) and to confirm linkage in previously identified chromosomal regions. DESIGN: Family-based linkage analysis. SETTING: Probands with familial LOAD identified in clinics in the Dominican Republic, Puerto Rico, and the United States. Patients We conducted a genome scan in 1161 members primarily clinically diagnosed as having LOAD; these members were from 209 families of Caribbean Hispanic ancestry. MAIN OUTCOME MEASURES: We analyzed 376 microsatellite markers with an average intermarker distance of 9.3 centimorgan. We conducted linkage analysis using possible and probable LOAD, and we performed affecteds-only 2-point linkage analyses assuming either an autosomal dominant or a recessive model. Subsequently, we conducted a multipoint affected sibling pair linkage analysis. RESULTS: Two-point parametric linkage analysis identified a locus at 3q28 with a genomewide empirical P value of .03 (logarithm of odds [LOD], 3.09) in a dominant model for probable and possible LOAD. Other regions suggestive of linkage included 2p25.3 (LOD, 1.77), 7p21.1 (LOD, 1.82), and 9q32 (LOD, 1.94). Under a recessive model, we also identified loci at 5p15.33 (LOD, 1.86), 12q24.21 (LOD, 2.43), 14q22.3 (LOD, 2.53), and 14q23.1 (LOD, 2.16) as suggestive for linkage. Restricted to probable LOAD, many of these loci continued to meet criteria suggestive for linkage, as did loci at 2p25.3 (LOD, 2.72), 3q28 (LOD, 2.28), 6p21.31 (LOD, 2.19), and 7p21.1 (LOD, 2.05). APOE conditional analysis indicated that the observed linkage at 3q28 was independent of the APOE epsilon4 allele. Multipoint nonparametric affected sibling pair linkage analysis provided confirmation of suggestive linkage for most, but not all, loci. CONCLUSIONS: Seven loci with LOD scores greater than 2.0 were identified among multiple affected Caribbean Hispanic families with LOAD. The highest LOD score was found at chromosome 3q28. At least 2 other independent studies have observed support for significant linkage at chromosome 3q28, highlighting this region as a locus for further genetic exploration.

Polymorphisms in the promoter of the human APP gene: functional evaluation and allele frequencies in Alzheimer disease.

BACKGROUND: Missense mutations in the amyloid precursor protein (APP) gene cause early-onset Alzheimer disease (AD). However, little is known regarding the effects of polymorphisms in regulatory sequences of APP on AD susceptibility. OBJECTIVES: To identify polymorphisms in the APP promoter, to test these for associations with AD, and to assess their influence on APP promoter activity in transfected cells. SETTING: Community study of 1013 people of white, African American, or Caribbean Hispanic ethnicity, 65 years and older, residing in northern Manhattan. MAIN OUTCOME MEASURES: The diagnosis of AD was established by stringent criteria, with multiple follow-up examinations over 7 years. RESULTS: We identified 2 polymorphisms in the APP promoter: a rare G-->C variant at -9 and a frequent G-->C variant at +37 relative to the transcription start site. The +37C allele was most frequent in African American patients (18% frequency), followed by Caribbean Hispanic patients (10%) and white patients of European descent (3%). This allele was overrepresented among patients with AD compared with elderly controls (odds ratio [OR], 1.57; 95% confidence interval [CI], 1.08-2.27 in the combined ethnic groups), but this was not significant after adjusting for age, sex, and education (OR, 1.41; 95% CI, 0.93-2.12). A stronger association was found in participants lacking any apolipoprotein-E epsilon4 allele (OR, 2.12; 95% CI, 1.36-3.32 [univariate analysis]; OR, 2.08; 95% CI, 1.26-3.45 after adjusting for age, sex, and education). The -9C allele was not frequent enough to be evaluated for a disease association. Both variants were tested in promoter-reporter assays in U-87 glioma cells, and no differences in promoter activity were detected. CONCLUSIONS: The -9G/C and +37G/C APP promoter polymorphisms are unlikely to contribute strongly to AD susceptibility or to cause major differences in APP expression, but the +37C allele warrants further study for association with AD in larger population samples.

APOE and APOC1 promoter polymorphisms and the risk of Alzheimer disease in African American and Caribbean Hispanic individuals.

BACKGROUND: The APOE epsilon4 allele is a genetic risk factor for Alzheimer disease (AD), though the strength of the association varies by ethnic group. Polymorphisms in regulatory sequences of APOE have also been related to AD, but the effects are inconsistent across studies. METHODS: We examined the association between AD and variants in 3 APOE promoters and in the promoter of the adjacent APOC1 gene in African American and Caribbean Hispanic individuals. Polymorphisms tested were the -491A/T, -427T/C, and -219G/T (Th1/E47cs) in the APOE promoter and the HpaI variant in the APOC1 promoter. Using standard research criteria for AD, overall odds ratios were computed and repeated stratified by presence or absence of APOE epsilon4. RESULTS: The APOC1 HpaI+ variant was associated with AD in Caribbean Hispanic individuals, but strong linkage disequilibrium with the APOE epsilon4 allele indicated that this was not an independent effect. No promoter variant in APOE or APOC1 was associated with AD before or after adjusting for age, education, sex, and multiple comparisons. Estimated haplotypes including -219G/T, APOE, and APOC1 differed significantly in Caribbean Hispanic patients and controls but not in African American participants. This effect was primarily owing to the -219G/T-APOE haplotype, but we did not detect significant allele-specific differences in promoter activity comparing reporter constructs containing the APOE -219G and -219 T alleles. CONCLUSION: These findings exclude a strong or independent influence of APOE or APOC1 promoter polymorphisms on the variation in APOE-related risk of AD in African American and Caribbean Hispanic individuals.