[MUSIC] In these presentations, we will examine some specific biases, Which are inherent to screening. By inherent to screening, I mean biases which are unavoidable while part of the screening process. If these biases are not dealt with, it would affect the estimates of screening, and indeed, it could lead to spuriously overestimate the benefit of screening. There are three biases we will look at. The first is a lead time bias. The second, the length time bias. And the third one, the overdiagnosis bias. So let's start with the lead time bias. As shown on the figures, without screening, at some stage, the patients will develop signs or symptom for disease which is going to be diagnosed. If a disease is detected at an advanced stage, the patient may actually die from the disease at some stage later. So the time interval between the detection, the diagnosis of the disease, and the death is called the survival. It's what you've got in red on the figures. No screening alters the natural history of the disease and actually does advance the time of diagnosis. By how much its advance will depend on the type of disease, the technology available on which the test is performed, and the time of advance diagnosis is called the lead time. The lead time is actually not observable in nature because the disease is detected earlier. We don't when it will have manifested clinically otherwise. Lead time is one beneficial effect from the screening. But per se, it is not sufficient to guarantee that screening is effective. If you look at the red row on the figures, despite the screening, if the person died at the same time, there will be no benefit from screening. Screening will be ineffective but survival will be artificially enhanced, because survival is calculated from the time of diagnosis until death. Therefore, we talk about bias of lead time because it actively improves survival, so it is not a good metric to estimate the impact of screening. Now, if screening is effective, the death is delayed, made from the same cause or for another unrelated cause, and we can measure the real impact of screening. The second type of bias is a Length time bias, which is usually found in disease with a very long pre-clinical phase. Every lesion has a different rate of progression, and screening tend to identify predominantly slow growing lesions. You can see on the figures, which is the white square, okay, this was a slowly progressing phase and a different screening episode the disease can be detected. Lesions that are rapidly progressive are harder to be picked up by screening. On the figures, too, are the four gray-shaped rectangles are lesions where have been picked up by these screening interventions. So when the phase, the pre-clinical phase, is shorter than the screening interval, the case cannot be detected by screening in most situations. The Length time bias is a selection bias based on the natural history of the disease. It is, as for the lead time, hard to quantify, but it can be observed in some instances. For instance, when you compare the stage distributions between screen-detected cases and cases with remiss at the screening episode, what we call for cancer, for instance, interval cancer can see that interval cancer tend to be more aggressive tumor with a poorer prognosis than the screen-detected lesions. So the length time is a bias because it leads to an artificial improvement in the stages of the disease when comparing screened and unscreened group. Therefore, the stage distribution is not a good indicator of the effectiveness of a screening program. Let's now look at the third bias, which is the overdiagnosis bias. But first, let's define overdiagnosis. As you can see on the lower part of the figure, overdiagnosis occurs when deaths, unrelated to the screened disease, occur before the clinical presentations of the disease might have occurred. What's interesting is that overdiagnosis definitions have changed over time, it used to be called pseudo-disease, a definition based only on the clinical and the pathological feature of the disease, as it was not related to the medical detection activity and screening, in particular. So overdiagnosis is actually a concept. It cannot be observed at either level, We do not know when a disease is detected, whether it is going to be another diagnosed case, or A more aggressive case of cancer or other detected disease. Overdiagnosis can be considered as extreme situations of lead time or length time bias. Overdiagnosis is an issue and the worse consequence of it is overtreatment of the disease, and it is a bias, because it increase the impact of screening. A little on disease who can not cause death will have 100% survival, so it will over-inflate the benefit from screening. So in conclusion, there are biases which are inherent to screening and therefore unavoidable. Any adequate evaluations of a screening program should take these biases into account by using appropriate methods about a choice of appropriate metrics. For instance, survival or state distributions of the disease comparing a screened and unscreened group are two measures which are highly susceptible to the screening biases. [MUSIC]