Definition[ edit ] The case—control is a type of epidemiological observational study. An observational study is a study in which subjects are not randomized to the exposed or unexposed groups, rather the subjects are observed in order to determine both their exposure and their outcome status and the exposure status is thus not determined by the researcher. Porta's Dictionary of Epidemiology defines the case—control study as: If a larger proportion of the cases smoke than the controls, that suggests, but does not conclusively show, that the hypothesis is valid.
W is correctly specified, are consistent if either of these models are correctly specified, and efficient if both are correctly specified.
Here A is the exposure of interest and If is a vector of covariates. We will compare the use of case-control weighted targeted maximum likelihood estimation in matched and unmatched case-control study designs as we explore which design yields the most information about the marginal causal effect.
This paper will not address matching in cohort studies, and will concentrate solely on case-control studies. However, matching in cohort studies was briefly addressed in van der Laanand applying our methods to cohort studies is an area of future research.
A Literature Review There is a large collection of literature devoted to the topic of individual matching in case-control study designs. This overview attempts to capture the most important consideratons, and it is by no means exhaustive.
Individual Matching in Case-Control Studies In an individually matched case-control study, the population of interest is identified, and cases are randomly sampled or selected based on particular inclusion criteria. Although, as Rothman and Greenland note, the definition of a case may implicitly define the population of interest for cases and controls.
Each of these cases is then matched to one or more controls based on a variable or variables believed to be a confounder. Much of the literature on individual matching in case-control studies, particularly earlier texts, describes these designs as a way to reduce confounding in the sampling design.
Reference to this is made in: MiettinenBreslow et al. However, several authors Breslow and Day, ; Kupper et al. Rothman and Greenland go on to say that while matching is intended to control confounding, it cannot do this in case-control study designs, and can, in fact, introduce bias.
Costanza agreed, stating that matching on confounders in case-control studies does nothing to remove the confounding, but frequently introduces negative confounding. So, while some literature cites the purpose of matching as improving validity, later publications Kupper et al.
Matched sampling leads to a balanced number of cases and controls across the levels of the selected matching variables. This balance can reduce the variance in the parameters of interest, which improves statistical efficiency.
A study with a randomly selected control group may yield some strata with an imbalance of cases and controls. It is important to add, however, that matching in case-control studies can lead to gains or losses in efficiency Kupper et al.
This will be discussed further in later sections. Breslow and Day note that matched case-control studies attempt to increase the informativeness of each of the subjects in the study.
However, one should also note that matched studies discard not only a pool of unmatched controls, but the information in each exposure-concordant case-control pair. Additionally, matching has a substantial impact on the study sample, most notably, it creates a sample of controls that is not representative of exposure in the population or the population as a whole.
The effect of the matching variable can no longer be studied directly, and the exposure frequency in the control sample will be shifted towards that of the cases Rothman and Greenland, Matching in case-control studies also does not completely control for the variable or variables used for matching, in general.
This means that researchers who implement matched designs must perform matched or stratified analyses Seigel and Greenhouse, ; Schlesselman, ; Holland and Rubin, ; Rothman and Greenland, ; Rubin, If an unmatched analysis is performed on matched data, the validity of the case-control comparison may be decreased Schlesselman, Variable Selection We revisit an earlier point made in this overview of individually matched case-control studies: If controls are matched to cases based on a variable that is not a true confounder, this can impact efficiency.
For example, if the matching variable is not associated with disease but is associated with the exposure, this will increase the variance of the estimator compared to an unmatched design.
Here, the matching leads to larger numbers of exposure-concordant case-control pairs, which are not informative in the analysis, leading to the increase in variance. If the matching variable is only associated with disease, there is often a loss of efficiency as well Schlesselman, If the matching variable is along the causal pathway between disease and exposure then matching will contribute bias that cannot be removed in the analysis Vandenbroucke et al.
Matching on a variable associated with exposure and not disease or a variable along the causal pathway are considered types of overmatching. Variables for matching should therefore be selected very carefully, and only those that are known to be associated with both exposure and disease should be considered.
The number of matching variables should also be reduced to as few as possible.
As the number of matching variables grows, the cases and controls will become increasingly similar with respect to the exposure of interest, and the study may produce a spurious result or provide no information Breslow and Day, Additionally, when matching on more than one variable, matching variables should not be strongly correlated with each other Schlesselman, More on Efficiency Kupper et al.Jan 01, · Why Match?
Investigating Matched Case-Control Study Designs with Causal Effect Estimation * We will compare the use of case-control weighted targeted maximum likelihood estimation in matched and unmatched case-control study designs as we explore which design yields the most information about the marginal causal effect.
This paper. Design: Unmatched case-control study. Setting: Principal maternity hospital of Kathmandu, Nepal. Subjects: All infants with neonatal encephalopathy from a population of 21 infants born over an 18 month period, and unmatched infants systematically recruited over 12 months.
Case definition and recruitment. Neonatal encephalopathy.
An unmatched case-control study was conducted using socio-demographic and behavioural data of individuals ( cases and controls) collected through an interviewer-administered questionnaire, comparing HIV patients late and early presented to care. Case-control study designs are used to estimate the relative risk for a disease from a specific risk factor.
The estimate is the odds ratio, which is a good estimate of the relative risk especially when the disease is rare. Case-control studies are useful when epidemiologists to investigate an. "Case-control studies are best understood by considering as the starting point a source population, which represents a hypothetical study population in which a cohort study might have been conducted.
The source population is the population that gives rise to the cases included in the study. A case–control study (also known as case–referent study) is a type of observational study in which two existing groups differing in outcome are identified and compared on .