"Confounding variables" here means "confounding bias," essentially. And this is true because a prospective cohort looks at a specific exposure to a substance (environmental toxin, drug, etc.), and asks "Who will develop this disease if exposed?" PCS's look attempt to find a relative risk associated with an exposure. They do not take into account the affects other exposures. This is your confounding bias. FA gives the example of confounding bias as "Pulmonary disease is more common in workers in a coal mine; however, miners are more likely to smoke," and since smoking can also lead to pulmonary disease, you can't really say whether the smoking (first or second hand) or the coal dust causes the problem. A clinical trial, on the other hand, contains a test group and a control group, so variables such as the confounding variable mentioned above are limited.