So now let's turn to the second of the major research designs, I'm going to talk to you about, and that is cohort studies. We're going to talk about cohort studies and case control studies. It turns out if you're working in this field just starting out, you're likely to do a case control study, before a cohort study. But cohort study is conceptually prior, for various reasons, so I'm putting that first. And here's the diagram, this is the standard diagram. A case control study and a cohort study differ, in their temporal orientation. So a cohort study has a, a putative exposure, something it thinks is dangerous or risky, and it compares people exposed, to people not exposed, and follows them forward in time. And over time, some people acquire the disease, that would be the cases and some people do not, that would be the controls. And so without too much difficulty, we can make a rate of incidence, or prevalence. We have, in the exposed group, a divided by a plus b, that would be the risk. And we can compare that risk, to the risk in the not exposed, that would be c, divided by C plus D, that would be the relative risk. And we can compare those two and if they're different, we're going to say, well that exposure is related to depressive disorder, let's say. It turns out, I'm showing you the relative odds also because that turns out to be crucial, for the case control study. You can see that, the quantity a over a plus b, is very similar to the quantity, a over b. That depends actually, how big a is, how many cases there are, and so we say with rare diseases, the odds ratio a over b, is very similar to the risk ratio, a over a plus b. So, if you're not a bookmaker, you're not used to relative odds, but epidemiologists talk about relative odds all the time. And we'll come back to this, when we talk about case control studies. But the point is, Cohort studies look forward in time and, as I'll show you, case control studies look backward in time. Now, the exemplar case control study, that I've taken from Lilienfeld and Stolley's book, is about smoking and lung cancer. We're going to have a case control study and a cohort study, about smoking and lung cancer. This cohort study, they enlisted 68,000 families and they got volunteers from the families, to inform about the people in the family. And actually, it concerned a million people, in those 68,000 families. And filled out questionnaires and they did annual follow ups. They lost some people out of, there were 27.000 people that they didn't, were not able to follow for one reason or other, but they got death certificates, and there were 46,000 people that died over a three year period, 1959 to 1962. And, here's what it shows, I think you all know this. Non-smokers have a lower rate of lung cancer. And people who smoked 40 plus cigarettes per day, had a very high rate of lung cancer. And, one of the interesting things is, the so called dosage effect. That is, you can see the more you smoke, the higher your risk for lung cancer. So that is a prospective study and that's pretty convincing, isn't it? That lung cancer is related to smoking. So here's a cohort study for depressive disorder and what we have to do is, we have to define the cohort event. So in the cohort study of smoking, it was just the start of the study. We had a population of about a million people. Here, we went to find an exposure that we think might be risky. And in this case, the exposure is, for women, a miscarriage. So we have 229 women who have had a miscarriage, and we follow them forward in time, for six months. And we find that 25 of them, have an occurrence of major depressive disorder. 25 over 229 is about 11%, 10.9%. and we have a community, controls, who are not exposed to miscarriage. These are women, who are of similar age and education and speak English, and they're matched on the season of the year also, so there are 230 of them and only 10 of them, have major depressive disorder, or 4.3%. So you can now see, we're now thinking, a miscarriage may be a risk factor for major depressive disorder and you can see, the relative risk: 25 over 229, that's about 10.9% over 10, divided by 230. That's about 4.3%, that generates a relative risk of 2.51. Now you have to bear with me a little bit because this, comes up again, in the case control study, the relative odds. Let's look at the relative odds, which are not 25 divided by 229, but 25 divided by 204. Right, and again, the odds of depression in the "not exposed" community group are 10 divided by not 230, but 220. and you can see that, Relative Odds is very very similar to the Relative Risk. Turns out to be crucial for the case control study, which comes up soon. So, the Cohort Study, the strengths of the Cohort Study is, one strength is, its perspective. So there's no, there's no, people can't, aren't remembering things incorrectly because you're proceeding forward in time. And you also can estimate incidence, which we discussed earlier, as the closest approximation to the force of morbidity. You can estimate incidence and because you can estimate incidence, you can estimate relative risk. But it turns out, the cohort study is hard to do, It's expensive. Often times in a cohort study, as in the smoking study, you have a small percentage of people exposed and many, many, many people who are not exposed. But you have to measure them anyway, and that costs money. And, if the disease takes a long time to occur, as in the case of lung cancer, you have to follow them for a long time. So that's a, a difficulty with the cohort study. And when you follow them for a long time, you will likely lose some people, as happened in the cohort study of smoking, where we lost 27,000 people. So, it could be that you'd lose the people, that are most likely, or perhaps least likely, to acquire the disease. And so we call that attrition bias. So, those are strengths and weaknesses of the Cohort Study. And now, let's go look at a different kind of study, very similar, but the case control study.