Marginal structural models for case-cohort study designs to estimate the association of antiretroviral therapy initiation with incident AIDS or death

To estimate the association of antiretroviral therapy initiation with incident acquired immunodeficiency syndrome (AIDS) or death while accounting for time-varying confounding in a cost-efficient manner, the authors combined a case-cohort study design with inverse probability-weighted estimation of a marginal structural Cox proportional hazards model. A total of 950 adults who were positive for human immunodeficiency virus type 1 were followed in 2 US cohort studies between 1995 and 2007. In the full cohort, 211 AIDS cases or deaths occurred during 4,456 person-years. In an illustrative 20% random subcohort of 190 participants, 41 AIDS cases or deaths occurred during 861 person-years. Accounting for measured confounders and determinants of dropout by inverse probability weighting, the full cohort hazard ratio was 0.41 (95% confidence interval: 0.26, 0.65) and the case-cohort hazard ratio was 0.47 (95% confidence interval: 0.26, 0.83). Standard multivariable-adjusted hazard ratios were closer to the null, regardless of study design. The precision lost with the case-cohort design was modest given the cost savings. Results from Monte Carlo simulations demonstrated that the proposed approach yields approximately unbiased estimates of the hazard ratio with appropriate confidence interval coverage. Marginal structural model analysis of case-cohort study designs provides a cost-efficient design coupled with an accurate analytic method for research settings in which there is time-varying confounding.     

Risk ratio and rate ratio estimation in case-cohort designs: Hypertension and cardiovascular mortality

Multivariate analysis in case-base designs depends on approximate methods. In the present study, new pseudo-likelihood methods are developed for this design. With these methods, the case-cohort risk ratio and rate ratio as well as their standard errors are easily estimated using logistic regression and Poisson regression, respectively. This is illustrated by the association between hypertension and cardiovascular mortality in a cohort, estimated by case-cohort analysis, using samples of several sizes. The estimates are compared with those obtained in full-cohort and nested case-control designs. The results indicate that these methods, which require nothing but widely available computer software, are valid. The case-cohort design, therefore, is a good, sometimes even advantageous alternative to the nested case-control design, in studying a disease that is not very rare. Application of the risk ratio method to the full cohort, using a ‘sample’ of 100 per cent follows logically; whenever the true risk ratio is desired instead of the odds ratio, a multivariate model for its estimation is therefore available.     

Analysis of case-cohort designs

The case-cohort design is most useful in analyzing time to failure in a large cohort in which failure is rare. Covariate information is collected from all failures and a representative sample of censored observations. Sampling is done without respect to time or disease status, and, therefore, the design is more flexible than a nested case-control design. Despite the efficiency of the methods, case-cohort designs are not often used because of perceived analytic complexity. In this article, we illustrate computation of a simple variance estimator and discuss model fitting techniques in SAS. Three different weighting methods are considered. Model fitting is demonstrated in an occupational exposure study of nickel refinery workers. The design is compared to a nested case-control design with respect to analysis and efficiency in a small simulation. In this example, case-cohort sampling from the full cohort was more efficient than using a comparable nested case-control design.     

Estimating interaction between genetic and environmental risk factors: efficiency of sampling designs within a cohort

Large prospective cohorts originally assembled to study environmental risk factors are increasingly exploited to study gene-environment interactions. Given the cost of genetic studies in large samples, being able to select a subsample for genotyping that contains most of the information from the cohort would lead to substantial savings. We consider nested case-control and case-cohort sampling designs with and without stratification and compare their efficiency relative to the entire cohort for estimating the effects of genetic and environmental risk factors and their interactions. Asymptotic calculations show that the relative efficiency of the case-cohort and nested case-control designs implementing the same sampling stratification are similar over a range of scenarios for the relationships among genes, environmental exposures, and disease status. Sampling equal numbers of exposed and unexposed subjects improves efficiency when the exposure is rare. The case-cohort designs had a slight advantage in simulations of sampling designs within the Framingham Offspring Study, using the interaction between apolipoprotein E and smoking on the risk of coronary heart disease as an example. It was possible to estimate the interaction effect with precision close to that of the full cohort when using case-cohort or nested case-control samples containing fewer than half the subjects of the cohort.     

Risk prediction measures for case-cohort and nested case-control designs: an application to cardiovascular disease

Case-cohort and nested case-control designs are often used to select an appropriate subsample of individuals from prospective cohort studies. Despite the great attention that has been given to the calculation of association estimators, no formal methods have been described for estimating risk prediction measures from these 2 sampling designs. Using real data from the Swedish Twin Registry (2004-2009), the authors sampled unstratified and stratified (matched) case-cohort and nested case-control subsamples and compared them with the full cohort (as "gold standard"). The real biomarker (high density lipoprotein cholesterol) and simulated biomarkers (BIO1 and BIO2) were studied in terms of association with cardiovascular disease, individual risk of cardiovascular disease at 3 years, and main prediction metrics. Overall, stratification improved efficiency, with stratified case-cohort designs being comparable to matched nested case-control designs. Individual risks and prediction measures calculated by using case-cohort and nested case-control designs after appropriate reweighting could be assessed with good efficiency, except for the finely matched nested case-control design, where matching variables could not be included in the individual risk estimation. In conclusion, the authors have shown that case-cohort and nested case-control designs can be used in settings where the research aim is to evaluate the prediction ability of new markers and that matching strategies for nested case-control designs may lead to biased prediction measures.     

External comparisons with the case-cohort design

The case-cohort design can be an economic alternative to the standard cohort design. Prentice (Biometrika 1986;73:1-11) showed how the case-cohort design can be used to obtain relative risk estimates for comparisons within the cohort being studied. In this paper, the authors consider ways in which the case-cohort design can be used for comparing risk in exposure groups within the cohort to the risk in an external population. The problem reduces to estimating the number of expected cases at each exposure level in the total cohort, when exposure status is available only for members of a subcohort, i.e., a random sample of the total cohort. The authors describe theoretical and empirical properties of several variations of the design and analysis of case-cohort studies. Empirical properties were examined by replicating the selection of the subcohort in a study of second cancer risk after chemotherapy for a first cancer. Use of a case-cohort design in that study would have saved five-sixths of the cost of gathering covariate information at the price of only an 11% loss in efficiency relative to a full cohort study.     

Case-cohort design in practice – experiences from the MORGAM Project

When carefully planned and analysed, the case-cohort design is a powerful choice for follow-up studies with multiple event types of interest. While the literature is rich with analysis methods for case-cohort data, little is written about the designing of a case-cohort study. Our experiences in designing, coordinating and analysing the MORGAM case-cohort study are potentially useful for other studies with similar characteristics. The motivation for using the case-cohort design in the MORGAM genetic study is discussed and issues relevant to its planning and analysis are studied. We propose solutions for appending the earlier case-cohort selection after an extension of the follow-up period and for achieving maximum overlap between earlier designs and the case-cohort design. Approaches for statistical analysis are studied in a simulation example based on the MORGAM data.     

Joint models for longitudinal and time-to-event data in a case-cohort design

Studies with longitudinal measurements are common in clinical research. Particular interest lies in studies where the repeated measurements are used to predict a time-to-event outcome, such as mortality, in a dynamic manner. If event rates in a study are low, however, and most information is to be expected from the patients experiencing the study endpoint, it may be more cost efficient to only use a subset of the data. One way of achieving this is by applying a case-cohort design, which selects all cases and only a random samples of the noncases. In the standard way of analyzing data in a case-cohort design, the noncases who were not selected are completely excluded from analysis; however, the overrepresentation of the cases will lead to bias. We propose to include survival information of all patients from the cohort in the analysis. We approach the fact that we do not have longitudinal information for a subset of the patients as a missing data problem and argue that the missingness mechanism is missing at random. Hence, results obtained from an appropriate model, such as a joint model, should remain valid. Simulations indicate that our method performs similar to fitting the model on a full cohort, both in terms of parameters estimates and predictions of survival probabilities. Estimating the model on the classical version of the case-cohort design shows clear bias and worse performance of the predictions. The procedure is further illustrated in data from a biomarker study on acute coronary syndrome patients, BIOMArCS.     

病例队列研究设计中相对危险度的估计及其应用

目的:系统介绍病例队列研究设计的基本原理,以及风险比（ HR）的常用估计方法及其应用。 方法:首先,介绍病例队列研究设计的基本原理;其次,对Prentice法、Self-Prentice法和Barlow法加权Cox比例风险回归模型进行描述和说明;最后,以上海市女性健康队列研究为例,分析全队列数据与...     

Piecewise exponential survival times and analysis of case-cohort data

Case-cohort designs select a random sample of a cohort to be used as control with cases arising from the follow-up of the cohort. Analyses of case-cohort studies with time-varying exposures that use Cox partial likelihood methods can be computer intensive. We propose a piecewise-exponential approach where Poisson regression model parameters are estimated from a pseudolikelihood and the corresponding variances are derived by applying Taylor linearization methods that are used in survey research. The proposed approach is evaluated using Monte Carlo simulations. An illustration is provided using data from the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study of male smokers in Finland, where a case-cohort study of serum glucose level and pancreatic cancer was analyzed.     

A full competing risk analysis of hospital-acquired infections can easily be performed by a case-cohort approach

ObjectivesWe provide a case-cohort approach and show that a full competing risk analysis is feasible even in a reduced data set. Competing events for hospital-acquired infections are death or discharge from the hospital because they preclude the observation of such infections.Study Design and SettingUsing surveillance data of 6,568 patient admissions (full cohort) from two Spanish intensive care units, we propose a case-cohort approach which uses only data from a random sample of the full cohort and all infected patients (the cases). We combine established methodology to study following measures: event-specific as well as subdistribution hazard ratios for all three events (infection, death, and discharge), cumulative hazards as well as incidence functions by risk factor, and also for all three events.ResultsCompared with the values from the full cohort, all measures are well approximated with the case-cohort design. For the event of interest (infection), event-specific and subdistribution hazards can be estimated with the full efficiency of the case-cohort design. So, standard errors are only slightly increased, whereas the precision of estimated hazards of the competing events is inflated according to the size of the subcohort.ConclusionThe case-cohort design provides an appropriate sampling design for studying hospital-acquired infections in a reduced data set. Potential effects of risk factors on the competing events (death and discharge) can be evaluated.     

Selecting appropriate study designs to address specific research questions in occupational epidemiology

Various epidemiological study designs are available to investigate illness and injury risks related to workplace exposures. The choice of study design to address a particular research question will be guided by the nature of the health outcome under study, its presumed relation to workplace exposures, and feasibility constraints. This review summarises the relative advantages and limitations of conventional study designs including cohort studies, cross-sectional studies, repeated measures studies, case-control (industry- and community-based) studies, and more recently developed variants of the nested case-control DESIGN: case-cohort and case-crossover studies.     

Designs and analysis of two-stage studies.

This paper concerns the design and analysis of two-stage studies, where, at the first stage, the response and the exposure variables are available among a large group of subjects. The other covariables, however, are available in only a subset of the large group, obtained in a second-stage sample. This paper introduces a class of twelve such two-stage designs, including two-stage case-control and case-cohort designs as special cases. In analysing such two-stage data, one objective is to extract information about the relationship between the exposure variable and the response after controlling for other covariables. We discuss three statistical methods to analyse the data and report results of Monte Carlo stimulation to study the efficiency of the three methods.     

Estimating the hazard rate difference from case-cohort studies

European Journal of Epidemiology - The case-cohort design, among many two-phase sampling designs, substantially reduces the cost of an epidemiological study by selecting more informative...     

Determination of the risk of a 2d cancer in patients treated for a first cancer

A paradoxical effect of radiotherapy and chemotherapy for cancer is that some of these treatments can themselves cause new cancers. Most epidemiologic methods can be applied successfully to the investigation of this problem and this paper reviews various approaches that have already been used by various researchers. The authors first review the more traditional methods, i.e., cohort and case-control studies and they then describe designs that have been proposed more recently, such as case-cohort studies. A distinction is established between internal comparisons, carried out within the study population, and external comparisons, in which a general population external to the population under study is used as the reference category. This presentation is mainly aimed at investigators using tumor registry data. However, the general principles formulated here are easily generalized to contexts other than that of registries.     

Nested case-control and case-cohort methods of sampling from a cohort: a critical comparison

The recently developed case-cohort method of sampling from a cohort is compared with the nested case-control method. Corrected asymptotic relative efficiency results show that the case-cohort design for single "disease" outcomes offers less improvement for intervention trials for which there is no random censoring than originally suggested. Furthermore, simulation results indicate that if there is moderate random censoring or staggered entry, the case-cohort method can do substantially worse than the nested case-control method.     

Premature death of adult adoptees: analyses of a case-cohort sample

Genetic and environmental influence on risk of premature death in adulthood was investigated by estimating the associations in total and cause-specific mortality of adult Danish adoptees and their biological and adoptive parents. Among all 14,425 non-familial adoptions formally granted in Denmark during the period 1924 through 1947, we selected the study population according to a case-cohort sampling design. As the case-control design, the case-cohort design has the advantage of economic data collection and little loss in statistical efficiency, but the case-cohort sample has the additional advantages that rate ratio estimates may be obtained, and re-use of the cohort sample in future studies of other outcomes is possible. Analyses were performed using Kalbfleisch and Lawless's estimator for hazard ratio, and robust estimation for variances. In the main analyses the sample was restricted to birth years of the adoptees 1924 and after, and age of transfer to the adoptive parents before 7 years, and age at death was restricted to 16 to 70 years. The results showed a higher mortality among adoptees, whose biological parents died in the age range of 16 to 70 years; this was significant for deaths from natural causes, vascular causes and all causes. No influence was seen from early death of adoptive parents, regardless of cause of death.     

Determining the effectiveness of the pentavalent rotavirus vaccine against rotavirus hospitalizations and emergency department visits using two study designs

The objective of this study is to determine the vaccine effectiveness (VE) of the pentavalent rotavirus vaccine (RV5) for preventing rotavirus-related hospitalizations and emergency department (ED) visits during the 2006–07 and 2007–08 rotavirus seasons using two study designs. Active, prospective population-based surveillance was conducted to identify cases of laboratory-confirmed rotavirus-related hospitalizations and ED visits to be used in case-cohort and case-control designs. VE was calculated using one comparison group for the case-cohort method and two comparison groups for the case-control method. The VE estimates produced by the three analyses were similar. Three doses of RV5 were effective for preventing rotavirus-related hospitalizations and ED visits in each analysis, with VE estimated as 92% in all three analyses. Two doses of RV5 were also effective, with VE ranging from 79% to 83%. A single dose was effective in the case-cohort analysis, but was not significant in either of the case-control analyses. The case-cohort and the case-control study designs produced the same VE point estimates for completion of the three dose series. Two and three doses of RV5 were effective in preventing rotavirus-related hospitalizations and ED visits.     

D-optimal cohort designs for linear mixed-effects models

The D-optimality criterion is used to construct optimal designs for different numbers of independent cohorts, which constitute a number of repeated measurements per subject over time. A cost function for longitudinal data is proposed, and the optimality criterion is optimized taking into account the cost of the study. First, an optimal number of design points for a given number of cohorts and cost was identified. Then, an optimal number of cohorts is identified by comparing the relative efficiencies (REs). A numerical study shows that for models describing the trend of a continuous outcome over time by polynomials, the most efficient number of repeated measurements is equal to the sum of the total number of cohorts and the degree of the polynomial in the model. REs of a purely longitudinal cohort design with only one cohort, and mixed longitudinal and cross-sectional cohort designs with more cohorts are compared. The results show that a purely longitudinal cohort design with only one cohort of subjects measured at the optimal time points is the most efficient design. The findings in this paper show that one can obtain a highly efficient design for parameter estimation with only a few repeated measurements. The results of this study will reduce the cost of data collection and ease the logistical burdens in cohort studies.