Monitoring dietary change in a low-fat diet intervention study: advantages of using 24-hour dietary recalls vs food records.

OBJECTIVE: The purpose of the study was to evaluate two methods of dietary assessment for monitoring change in fat intake in a low-fat diet intervention study. DESIGN: The two dietary assessment methods were a 4-day food record (4DFR) and an unannounced 24-hour dietary recall conducted by telephone interview (referred to as a telephone recall [TR]). Subjects were assigned randomly to either a low-fat diet intervention group or a control group that received no counseling about fat intake. Dietary data were collected at baseline, 6 months, and 12 months. SUBJECTS: Two hundred ninety postmenopausal women with localized breast cancer were recruited at seven clinical centers in the United States. STATISTICAL ANALYSIS: Analysis of variance was used to test for significant differences in mean fat and energy intakes. RESULTS: Three sources of error were identified: (a) an instrument effect, suggesting underreporting at baseline of approximately 8% in mean energy intake and 11% in mean fat intake in the TR group compared with the 4DFR group (P = .0001); (b) a repeated measures effect observed for the 4DFR, suggesting underreporting of approximately 7% for energy intake and 14% for fat intake in the control group at 6 and 12 months compared with baseline values (P < .001); and (c) an adherence effect (or compliance bias), suggesting greater compliance to the low-fat intervention diet when subjects were keeping food records than when estimates were based on the unannounced TR. Compared with the TR, the 4DFR overestimated the extent of fat reduction in the low-fat diet intervention group by 41% (P = .08) and 25% (P = .62) at 6 and 12 months, respectively. APPLICATION: Multiple days of unannounced 24-hour recalls may be preferable to multiple-day food records for monitoring dietary change in diet intervention studies.

Detection of gene-environment interactions in joint segregation and linkage analysis.

We compare approaches for analysis of gene-environment (G x E) interaction, using segregation and joint segregation and linkage analyses of a quantitative trait. Analyses of triglyceride levels in a single large pedigree demonstrate the two methods and show evidence for a significant interaction (P=.015 when segregation analysis is used; P=.006 when joint analysis is used) between a codominant major gene and body-mass index. Genotype-specific correlation coefficients, between triglyceride levels and body-mass index, estimated from the joint model are rAA=.72, rAa=.49, and raa=. 20. Several simulation studies indicate that joint segregation and linkage analysis leads to less-biased and more-efficient estimates of a G x E-interaction effect, compared with segregation analysis alone. Depending on the heterozygosity of the marker locus and its proximity to the trait locus, we found joint analysis to be as much as 70% more efficient than segregation analysis, for estimation of a G x E-interaction effect. Over a variety of parameter combinations, joint analysis also led to moderate (5%-10%) increases in power to detect the interaction. On the basis of these results, we suggest the use of combined segregation and linkage analysis for improved estimation of G x E-interaction effects when the underlying trait gene is unmeasured.

Simultaneously modelling censored survival data and repeatedly measured covariates: a Gibbs sampling approach.

Recent methodologic developments in the analysis of longitudinal data have typically addressed one of two aspects: (i) the modelling of repeated measurements of a covariate as a function of time or other covariates, or (ii) the modelling of the effect of a covariate on disease risk. In this paper, we address both of these issues in a single analysis by modelling a continuous covariate over time and simultaneously relating the covariate to disease risk. We use the Markov chain Monte Carlo technique of Gibbs sampling to estimate the joint posterior distribution of the unknown parameters of the model. Simulation studies showed that jointly modelling survival and covariate data reduced bias in parameter estimates due to covariate measurement error and informative censoring. We illustrate the methodology by application to a data set that consists of repeated measurements of the immunologic marker CD4 and times of diagnosis of AIDS for a cohort of anti-HIV-1 positive recipients of anti-HIV-1 positive blood transfusions. We assume a linear random effects model with subject-specific intercepts and slopes and normal errors for the true log and square root CD4 counts, and a proportional hazards model for AIDS-free survival time expressed as a function of current true CD4 value. On the square root scale, the joint approach yielded a mean slope for CD4 that was 7 per cent steeper and a log relative risk of AIDS that was 35 per cent larger than those obtained by analysis of the component sub-models separately.

Joint segregation and linkage analysis of a quantitative trait compared to separate analyses.

Our goal was to determine the degree to which joint segregation and linkage analysis leads to increased efficiency for estimating the recombination fraction and to greater power for detecting linkage, compared to separate analyses. We concentrated on the quantitative phenotype Q2 and analyzed linkage with a tightly linked marker, a loosely linked marker, and eight unlinked markers, the latter chosen to evaluate false positive rates. We considered both nuclear-family and extended-pedigree data, using the 200 replicates of each provided to GAW participants. We found joint analysis to be consistently more efficient, with relative efficiencies for the tightly linked marker of 1.16 and 1.06 in extended pedigrees and nuclear families, respectively. These relative efficiencies translated into modest but consistent gains in power to detect linkage. Both methods appear to produce unbiased parameter estimates and similar false positive rates.

Genetic epidemiologic analysis of quantitative phenotypes using Gibbs sampling.

We analyzed two quantitative traits (Q1 and Q2) provided in the 'Common Disease' data set with the aim of detecting both genetic and environmental determinants. We used linear regression for screening measured variables, maximum likelihood segregation and linkage analyses for detecting and localizing unmeasured genes, and Gibbs sampling for joint segregation and linkage analyses with estimation of gene-environment interaction and polygenic effects. For both Q1 and Q2, we successfully detected the unmeasured codominant major gene (MG) that was tightly linked to candidate gene C2. We also detected all of the measured variables used in generating Q1 (age, Q3, candidate gene C5) and Q2 (EF). Although our final models differed slightly from the true data generation models, our multifaceted analytic approach was successful in characterizing the determinants of Q1 and Q2.