The 2x2 table shown in Figure 7-4 is generic — meaning it can be filled in with data from a

cross-sectional study, a case-control study, a cohort study, or even a clinical trial (if you replace

the E+ and E– entries with intervention group assignment). How the results are interpreted from

the 2x2 table depend upon the underlying study design. In the case of a cross-sectional study, an

odds ratio (OR) could be calculated to quantify the strength of association between the exposure

and outcome (see Chapter 14). However, any results coming from a 2x2 table do not control for

confounding, which is a bias introduced by a nuisance variable associated with the exposure and

the outcome, but not on the causal pathway between the exposure and outcome (more on

confounding in Chapter 20).

Imagine that you were examining the cross-sectional association between having the exposure

of obesity (yes/no), and having the outcome of HTN (yes/no). Household income may be a

confounding variable, because lower income levels are associated with barriers to access to

high-quality nutrition that could prevent both obesity and HTN. However, in a bivariate analysis

like is done in a 2x2 table, there is no ability to control for confounding. To do that, you need to

use a regression model like the ones described in Chapters 15 through 23.

So how would you use a 2x2 table for a case-control study on a statistically rare condition like liver

cancer? Suppose that patients thought to have liver cancer are referred to a cancer center to undergo

biopsies. Those with biopsies that are positive for liver cancer are placed in a registry. Suppose that

in 2023 there were 30 cases of liver cancer found at this center that were placed in the registry. This

would be a case series. Imagine that you had a hypothesis — that high levels of alcohol intake may

have caused the liver cancer. You could interview the cases to determine their exposure status, or level

of alcohol intake before they were diagnosed with liver cancer. Imagine that 10 of the 30 reported high

alcohol intake. You will see that as some evidence for your hypothesis.

But you could not do causal inference unless you had a comparable comparison group without liver

cancer so that you could fill out your 2x2 table. Imagine that you went back to the cancer center and

were able to contact and enroll 30 patients who had liver biopsies but were found not to have liver

cancer to serve as controls. Suppose that you interviewed this group and discovered that only two of

them reported high levels of alcohol intake. You could develop the 2x2 table like the one shown in

Figure 7-5.

© John Wiley & Sons, Inc.