Opting for a different ratio than 1:1 is done uniquely for a single reason: finding participants for one of the arms (e.i. cases - control or exposed - non-exposed) is difficult whereas it seems much easier for the other. For the same statistics power to detect a difference, this will however increase the total number of needed participants. The power gained for ratios above 1:3 is poor compared to the additional workload it requires. Therefore it is rare to see ratios above 1:3.
Ratio is not to confuse with matching. You can compare groups of unequal size or match participants together in more than a pair. Matching is designed to control for unmeasured confounders. It is therefore very important to only match on factors known to be related to the condition of interest. Furthermore, matching requires using different statistical analysis approach than non-matched samples.
I therefore suggest the following rules:
When possible always use a 1:1 ratio unless data is already collected.
Favor collecting information on known potential confounders and adjust at the statistical analysis stage rather than correcting at the design stage using matching.
Matching increases the efficiency of the estimates if the matching variables are associated with both the disease and the exposure. Barring the cost and time limitations of including additional controls, there are two reasons for considering higher matching schemes (ie. 1:2, 1:3..etc): 1) concern for sufficient numbers in stratified analysis; 2) the increase in power given the expected prevalence of exposure among the controls (see Hennessy et al. "Factors Influencing the Optimal Control-to-Case Ratio in Matched Case-Control Studies," American Journal of Epid 1999; 149(2))
a higher ratio than 1:1 for controls:cases incresaes statistical power, but this effect is negligible above 4:1. cost considerations also can become important.
Your question: What is the Rationale for 1:2 ratio in Case Control studies?
In Case Control studies, the selection of cases and controls in the ratio of 1:1 then only your result precision would increase and you would get the current results for your particular study. Your study power will also be increase. So, very careful at the time of selection of sample for this type of studies.
In Case-Control studies, three sources of bias are occur.
Your first question was deeply answered. Regarding the second one (how), it depends of your specific needs, specially when controlling for confounding.
The first choice is matching, which can be done for one, two or several variables. Advanced matching includes techniques like propensity scores. The most common matching is age and gender, where each case is matched with a control with the same qualities. In case of several potencial matches, the choice is made randomly. Sometimes an exact match cannot be found, in which case the closest one is chosen.
Although age and sex are traditional variables for matching, it is not necessary using them unless you suspect that those are confusion variables. Any other variable can be chosen for matching as long as it is considered an important confounder. Please remember that matched data require specific data analysis (paired tests) in order to account for the matching.
If you are nor interested in controlling confounders by matching, the ratio of 2 controls for case can be easily achieved by picking a random sample among your selected controls. Be careful of selection bias, as the rules for control selection also apply in this situation.
Opting for a different ratio than 1:1 is done uniquely for a single reason: finding participants for one of the arms (e.i. cases - control or exposed - non-exposed) is difficult whereas it seems much easier for the other. For the same statistics power to detect a difference, this will however increase the total number of needed participants. The power gained for ratios above 1:3 is poor compared to the additional workload it requires. Therefore it is rare to see ratios above 1:3.
Ratio is not to confuse with matching. You can compare groups of unequal size or match participants together in more than a pair. Matching is designed to control for unmeasured confounders. It is therefore very important to only match on factors known to be related to the condition of interest. Furthermore, matching requires using different statistical analysis approach than non-matched samples.
I therefore suggest the following rules:
When possible always use a 1:1 ratio unless data is already collected.
Favor collecting information on known potential confounders and adjust at the statistical analysis stage rather than correcting at the design stage using matching.