Matching in Epidemiologic Studies: Validity and Efficiency Considerations

Kupper, Lawrence L.; Karon, John M.; Kleinbaum, David G.; Morgenstern, Hal; Lewis, Donald K.

doi:10.2307/2530417

Cited by 101 publications

(63 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Where OR Ob-PCOS is weak to moderate (i.e., <5.0) matching elicits no benefit (i.e., β 0 for the red plot) or even a slight detriment (i.e., β <0 for the blue plot) in OR PCOS-IR precision compared with independent control selection. These observations are consistent with patterns reported from statistical simulations of matching strategies (16,18).…”

supporting

confidence: 91%

“…Using both independent and matching control selection strategies 100 each PCOS cases and controls are considered with respect to insulin resistance as a hypothesized causal risk factor (14) and obesity (15) as a potential confounding variable. These date were generated using previously published formulae for expected values under independent and matched control selection strategies (16), with Excel 2003 (Microsoft Co., Redmond, WA), by varying the magnitude of the odds ratios between obesity and PCOS (OR Ob-PCOS ) from 1.0 (i.e., no association) to 12.0 (very strong association) and obesity and insulin resistance (OR Ob-IR ) from 1.0 to 7.4. Count data were subsequently analyzed with SAS v. 9.0 (SAS Institute, Inc., Cary, NC) to generate obesity adjusted odds ratios (17) and 95% confidence intervals for insulin resistance as a risk factor for PCOS (OR PCOS-IR ).…”

mentioning

confidence: 99%

See 1 more Smart Citation

The use and misuse of matching in case-control studies: the example of polycystic ovary syndrome

Bloom

Hediger

2007

Fertility and Sterility

View full text Add to dashboard Cite

Matching control selection strategies are often employed in PCOS case-control studies; however, they are infrequently used in an appropriate fashion. When properly applied, matching may offer improved study precision, but this is highly contingent on the causal pathway under consideration, strength of the associations between the matching variable and both the risk factor of interest and PCOS, and use of an appropriate stratified data analysis.Variations in design, including strategies to consider suspected confounding variables, may be in part responsible for discrepancies among study results reported in the polycystic ovary syndrome (PCOS) literature (1). The case-control study design is frequently used to consider hypothesized risk factors for PCOS, with investigators often complementing this framework with a 'matching' strategy in which controls are selected for cases according to the distribution of suspected confounding variables among the latter (2). A recent PubMed search (i.e., on 6/14/06), using the search terms 'polycystic ovary syndrome' and 'case-control studies', limited to the English language and 'published in the last 1 year', generated 40 citations, 23 of which fell under the PCOS case-control rubric. Almost half of these 23 studies, 11 (48%), matched controls to cases by body mass index (BMI), age, or both, but only one (9%) analyzed the data appropriately for its matched nature. In a prior publication, we discussed the merits of the case-control design for studies of PCOS causal risk factors (3); however, the proper

show abstract

supporting

confidence: 91%

mentioning

confidence: 99%

The use and misuse of matching in case-control studies: the example of polycystic ovary syndrome

Bloom

Hediger

2007

Fertility and Sterility

View full text Add to dashboard Cite

show abstract

“…Others have approached the problem from an experimental viewpoint, comparing the efficiencies of unmatched versus matched designs. For an overview of some of the results, see Rubin (1973), McKinlay (1977), Kupper et al (1981), andGreenland (1986).…”

Section: Introductionmentioning

confidence: 99%

When Does It Pay to Break the Matches for Analysis of a Matched-Pairs Design?

Lynn¹,

McCulloch²

1992

Biometrics

View full text Add to dashboard Cite

SUMMARYTwo methods of analyses are compared to estimate the treatment effect of a comparative study where each treated individual is matched with a single control at the design stage. The usual matched pairs analysis accounts for the pairing directly in its model, whereas regression adjustment ignores the matching but instead models the pairing using a set of covariates. For a normal linear model, the estimated treatment effect from the matched pairs analysis (paired t-test) is more efficient. For a Bernoulli logistic model, matched pairs analysis performs better when the sample size was small, but is inferior to logistic regression for large sample sizes.

show abstract

“…Reference to this is made in: Miettinen (1970), Breslow et al (1978), Breslow and Day (1980), Kupper et al (1981), Schlesselman (1982), Collett (1991), andCostanza (1995), among others. However, several authors (Breslow and Day, 1980;Kupper et al, 1981;Schlesselman, 1982;Rothman and Greenland, 1998;Vandenbroucke et al, 2007) point out that the goal of matching is to increase the study's efficiency by forcing the case and control samples to have similar distributions across confounding variables. Rothman and Greenland (1998) 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.…”

Section: Individual Matching In Case-control Studiesmentioning

confidence: 99%

Why Match? Investigating Matched Case-Control Study Designs with Causal Effect Estimation

Rose¹,

Laan²

2009

The International Journal of Biostatistics

253

190

View full text Add to dashboard Cite

Matched case-control study designs are commonly implemented in the field of public health. While matching is intended to eliminate confounding, the main potential benefit of matching in case-control studies is a gain in efficiency. Methods for analyzing matched case-control studies have focused on utilizing conditional logistic regression models that provide conditional and not causal estimates of the odds ratio. This article investigates the use of case-control weighted targeted maximum likelihood estimation to obtain marginal causal effects in matched case-control study designs. We compare the use of case-control weighted targeted maximum likelihood estimation in matched and unmatched designs in an effort to explore which design yields the most information about the marginal causal effect. The procedures require knowledge of certain prevalence probabilities and were previously described by van der Laan (2008). In many practical situations where a causal effect is the parameter of interest, researchers may be better served using an unmatched design.

show abstract

Matching in Epidemiologic Studies: Validity and Efficiency Considerations

Cited by 101 publications

References 23 publications

The use and misuse of matching in case-control studies: the example of polycystic ovary syndrome

The use and misuse of matching in case-control studies: the example of polycystic ovary syndrome

When Does It Pay to Break the Matches for Analysis of a Matched-Pairs Design?

Why Match? Investigating Matched Case-Control Study Designs with Causal Effect Estimation

Contact Info

Product

Resources

About