A research group is interested in the association between vitamin E supplementation and a decreased risk for osteoporosis in a group of elderly women. The team randomly samples a group of women potentially enrolled in the study after a strict set of inclusion criteria have been met, and pairs each woman receiving the supplementation with two women receiving a placebo based on a number of characteristics such as BMI, education, and socio-economic status. The data compiled after 10 years of follow up were analyzed using conditional logistic regression. Which of the following methods to reduce confounding was not employed in the above scenario?

1. Stratification
2. Regression
3. Restriction
4. Matching