Skip to 0 minutes and 15 secondsWhenever we encounter a cancer, there are some fairly obvious questions that come to mind. How common is it? How does it vary? Who gets it? Is it becoming more or less common? What's causing it? Can we treat it? What are the outcomes? How do we detect it? These are questions which can be answered by the science of epidemiology. I want to say a word or two about just how epidemiology can answer these questions for us. Basically, we divide epidemiology into two parts, descriptive epidemiology and analytical epidemiology. Descriptive epidemiology answers two broad questions. The first is, how common. And the second is, how does the tumour vary by person, place, and time.
Skip to 0 minutes and 59 secondsWhen it comes to how common, we use incidence and prevalence. And together these give us a very clear idea of events, like the onset of cancer, and states, like living with cancer. So for instance, incidence is the number of new cases in a defined population over a length of time. So let's consider cervical cancer. The cancer registry, that's often how we collect these data - each country has a registry that captures events and analyses them and brings information together. The registry counts that, for example, there were about 1,000 cases cervical cancer this population. But that information on its own is not very helpful. It's what we call a floating numerator.
Skip to 1 minute and 48 secondsWe need a denominator which is the number of people who were at risk of developing that cancer in that year. And for that, we need to think, who is at risk? Well in the case of cervical cancer, it would be females not males. And given what we know about the causation of cervical cancer, we'd probably say, females above the reproductive age. And so we have the incidence rate, the number of new cases divided by the number of people at risk over a time period, usually a year. But if we're interested in the state of living with cancer as opposed to events like developing cancer and dying from cancer, we use the prevalence ratio.
Skip to 2 minutes and 30 secondsAnd for that, we have the number of people with the cancer divided by those who were risk of it. And so we have a ratio of those with cancer, the state of cancer, divided by those at risk. Then once we've got our incidence and our prevalence data, we ask ourselves, well, how does it vary? Who is suffering from this cancer? Is it becoming more or less common? And where are these people? So for example, in the case of smoking and lung cancer before we knew that smoking was the cause of lung cancer, in the 20th Century, we observe that males in industrialised countries were developing more lung cancer. And that begged the question, why?
Skip to 3 minutes and 12 secondsAnd then soon afterwards, the female rate started creeping up. And various hypotheses were would put forward as to why this might be. Equally, the variability by place has got a lovely example of Burkitt's lymphoma. Burkitt was a surgeon working in Africa. And he began to see these tumours in young African children and he went on what he called a tumour safari. And he went round and they mapped where the tumour could be found. And he also mapped where malaria was prevalent. And he found a very clear correspondence. And that gave rise to the awareness that it was the compromising of immunity that gave rise to the predisposition to the Burkitt's lymphoma cancer.
Skip to 3 minutes and 53 secondsSo these person, place, and time incidence prevalence data give us clues, help us to generate hypotheses, that can be tested by analytical epidemiology. And the two big tools of analytical epidemiology are case control studies and cohort studies. And we'll go back to our example of smoking. In the 1950s, Richard Doll and others began to suspect that smoking might well be the cause of lung cancer. So he conducted a series of case control studies. He took people with lung cancer, and he matched them with controls, people who were as similar as possible but without the cancer. He then inquired about their back story, their histories.
Skip to 4 minutes and 35 secondsAnd he found, of course, that there were many more smokers amongst those with cancer than those amongst the controls. That, however, didn't prove causation. It's possible that it was the onset of the cancer or the symptoms of it that caused people to take up smoking. We know that's not the case. But that's the kind of bias that can creep into case control studies. So what you do now is, you conduct a cohort study. You take a whole population. You itemise the risk factors in that population. And you follow them leading into the future. And you detect those people who develop cancers as they do so. And that methodology is very, very powerful in establishing causation.
Skip to 5 minutes and 20 secondsBecause if you get something that was in existence before the cancer and, particularly, if you see that there's a big difference in the proportion of people with the risk who develop the cancer against those who don't. And also you see dose response. More smoking. More lung cancer. These factors come together in a set of criteria that help us establish causation. Once we know what's causing the cancer, we can begin to think about detection and treatment. And detection is usually done by formal screening or at least by case finding by clinicians and others when patients encounter to them.
Skip to 6 minutes and 0 secondsSo for example, there's a simple test that we can apply to look for blood in the bowel motions which suggests the possibility of colon cancer. And that test has been used in a population-wide basis to detect colon cancer early so that it can be more be easily removed, indeed, curatively removed by surgery. That's a great boon. But there's dangers in that. If we simply detect the cancer early and cause someone to live for a longer period of time but no longer than they might otherwise have done so, that's called lead time bias. And equally so, take prostate cancer. More people die with prostate cancer than from prostate cancer.
Skip to 6 minutes and 46 secondsSo simply detecting a cancer which wouldn't have done the patient any harm might not be doing them any favours. So there are dangers and issues that epidemiology can help resolve in the whole business of screening and detection. Then finally, there is the issue of treatment. The efficacy of treatment is best established by randomised controlled trials. And in the case of cancer, we usually ally that to what we call survival curves. So each arm of the trial has a survival curve and we can see what proportion of those who are treated are still alive at one, three, five, and more years. And that allows us to make accurate comparisons of the efficacy of the treatment being considered.
Skip to 7 minutes and 31 secondsSo in summary, this is what epidemiology does to help us understand cancers. It establishes how common, through incidence and prevalence. It establishes variability in terms of person, place, and time. It helps us explore causation through case control and cohort studies. It helps us organise detection and screening. And finally it establishes the efficacy of rival treatments.
Professor Phil Hanlon describes the importance of epidemiology in studying cancer, and explains some key epidemiological concepts.
© University of Glasgow