I'm Doug Arenberg, Associate Professor of Medicine at the University of Michigan Medical School in the division of pulmonary and critical care medicine. My clinical practice and interests are focused on lung cancer, and in this lecture I'm going to be talking about lung cancer screening. The objectives of this lecture are to come away with an understanding of the history and the rationale for current lung cancer screening guidelines, and in that recognize the importance of our ability to measure lung cancer risk. And with the patient, be able to discuss the risks and benefits of screening with people who are at risk for lung cancer. So we'll start with this question, and I want you to think of a hypothetical screening test. Which of the following findings would support the use of this hypothetical screening test? So A, a screening test which leads to improved 5 year survival of screened patients from 30% in the control group to 80% in the screened group. B, a cancer screening blood test which detects more cases of cancer it is intended to find. C, a cancer screening test which finds more cases of early stage disease when compared with a control group. Or D, a test for which the mortality from the screened disease is 1.6% in the screened group and 1.9% in the control group. So we're going to start with the historical perspective, and for some 30 years, cancer screening recommendations were based on statements made by the American Cancer Society as well as the US Preventive Services Task Force. And to put this into context, the US Preventive Services Task Force is a group of volunteers, experts in healthcare who issue recommendations for screening services and preventive services. Everything from how to check blood pressure, to when to vaccinate, to how and whether to screen for cancer. In 1980, they issued a statement, both the American Cancer Society and the US Preventive Services Task Force issued recommendations that said chest X-rays were not indicated for the early detection of lung cancer. And that people with signs and symptoms of lung cancer should consult their physician. Prior to 1980, it was just assumed that chest X-rays were a good idea for smokers. But as I'll show you later, several studies done in the 1970s showed that chest X-rays did not serve as a good screening test for lung cancer and therefore should not be used for that purpose. In that context, deaths in the United States alone from lung cancer increased from about 117,000 in 1980 to over 200,000 in 2006. So where are we now? Well, recently the US Preventive Services Task Force reevaluated the available data and came up with this recommendation, published in the Annals of Internal Medicine in 2014. And what they said was that the US Preventive Services Task Force recommends annual screening for lung cancer with low-dose CT scan in adults aged 55 to 80 who had a 30 pack-year smoking history within the last 15 years if they were quitters. So you either had to be a current or former smoker within the last 15 years, having smoked 30 pack-years or more, and between the ages of 55 and 80. And they also recommended that the screening be discontinued once a person has quit for 15 years or more or, and this is important, develops a health problem that substantially limits life expectancy or the ability or willingness to have curative lung surgery. So where did this recommendation come from? It came from one of the largest studies ever conducted by the National Institutes of Health, the National Lung Screening Trial, and this study was an enormous undertaking. They enrolled 50,000 healthy, and I emphasize healthy, current or former smokers and again, former smokers had to be quitters within the last 15 years. They had to be heavy smokers defined as 30 pack-years or more. And a pack-year is simply 1 pack per day for a year would be 1 pack-year, 2 packs per day for 15 years would be 30 pack-years and so on, and so forth. So 50,000 healthy current or former smokers between the ages of 55 and 74. And they were randomized to either get annual chest X-ray or annual low dose CT scan at base line, then again one year and two years later. The study was powered to detect a 20% mortality benefit. Primary outcomes also included all cause mortality, the prevalence, incidence, the detection of interval cancers, the positive predictive value, the negative predictive value, and the stage distribution of lung cancer. Importantly, they also measured quality of life of those involved in the study and even anxiety related to screening tests. This is one of the most well-done undertaken screening studies ever in history. And they also measured medical resource utilization for positive screen, so health economists could have a go with the data when the study was completed. And this is what they found. This curve shows, on the upper curve, the number of cumulative deaths from lung cancer and those who were randomized to chest X-rays. In the lower curve, the black curve, the cumulative number of lung cancer deaths amongst those who were randomized to low-dose CT screening. And what you see is that over time, the accumulation of lung cancer death is far greater in the control group than it is in the low-dose CT group, a 20% reduction in lung cancer mortality, exactly what the designers of the study anticipated finding. And this doesn't sound like much, but when you compare it to other screening tests, a 20% mortality benefit is at least as good as, if not better than, every other currently employed cancer screening test that we do. So the screening was done just in these early years time 0, time 1, and time 2 years, and then look what happened to these two groups as time passed to the number of cancer deaths. And what you see is the accumulation of excess lung cancer deaths in the control group relative to the low-dose CT group. So if you develop a score card from this data, what you get is a balance of benefits and harms, and we should review these in some detail. So the benefit was obvious. If you got screened with low-dose CT, you were 20% less likely to die from lung cancer than if you were screened in the controlled group with chest X-rays. And I should add that we now know that chest X-rays don't reduce lung cancer mortality at all. So essentially, they served as a control group. But to get this 20% mortality benefit, you pay a price. And the biggest price, which is one we'll talk about later, is that a lot of people have lung nodules, and that's what we're looking for in these CT scans. About 36% of participants in the National Lung Screening Trial had a lung nodule that was a false positive. And in some of these people, 2.5% to be precise, they ended up having a false alarm that led to some invasive procedure, a biopsy of some sort, a bronchoscopy, or even surgery. So that's 25 out of 1,000 people, and 3 out of 1,000 had some major complication as a result of these invasive procedures. So you pay this price to get that benefit. And the biggest price is driven by the finding of false positives in this group. So colleagues of mine did a study to ask what happens if you look at the entire population of people that were screened in this group? Did they all have the same experience? And this is the table From a paper that they published to show that the experience was very different amongst persons of different risk for cancer. And this paper was actually prompted by and advertisement that they saw in a newspaper, out in California, advertising in the bottom row there, the Sequoia Hospital Lung Cancer Screening Program. Now, not to single them out, but the Sequoia Hospital thought that they should be offering screening to everyone, based on this data, who had never really even come close to having risk for lung cancer. Including, if you're 40 years old and had smoked perhaps just as little as ten pack years. The absolute risk of lung cancer in that group was quite low, and therefore, the number of deaths that could be prevented in that group was also quite low. And what they found was that to save one life, they would have to screen 35,000 people in that risk group. Compare that with the top three rows. The average participant in the National Lung Screening Trial, that is person who is about in the mid-range in terms of their risk, a 62-year-old male, current smoker, about a pack and-a-half a day smoker for thirty five years. The absolute number of deaths per thousand people in that group is close to twenty. If you reduce that by 20%, you get 15.6. So you avert nearly 4 deaths per 1,000 persons, and it's a screen to save 1 life, you could screen 256 people. This is, believe it or not, a very, very efficient screening strategy. Well, let's take the person who is minimally eligible for the National Lung Screening Trial. A 55 year old female, former smoker, a pack per day for 30 years, who quit. That person, if you take a 1,000 people just like them they're absolute risk is about 4 of those 1,000 are going to die from lung cancer. Reducing that by 20% you get 3.2. You take the inverse of that, you get 1200 people that you need to screen to save 1 life from lung cancer. And so on and so forth. If you take people who are very high risk, it's a very efficient screening strategy. People who are very low risk, it becomes very inefficient, very costly and we cannot ignore the risks because the risks are constant across that, I should say, the harms of screening are constant across that risk spectrum. So we really want to make sure that we focus this service on people who are at the highest risk for lung cancer. These are the people who can benefit the most from the test. So let's look at it from another perspective. From a population perspective, let's take 1,000 high risk participants who would have been eligible for the National Lung Screening Trial. We get 3 fewer deaths by screening them, that's good. On the other side, there's the false positives, 365 throughout 36% as we said. 25 biopsies, a complication called overdiagnosis, which we can talk about a little bit more. And 3 major complications from invasive procedures, 6 out of 10,000, 0.6 out of 1,000 will die as a result of being screened, not because they had cancer. So you save 3 lives, but you may put somebody in danger as a result of saving those 3 lives out of a 1,000 people. On balance, that's a pretty good outcome, particularly if you're one of those three people whose life was saved. Now, what if they're a lower risk group? Take 1,000 people who are very low risk. There you don't get as many life save. In other words, you'd have to screen 10,000 to save those 3 lives, but for every 1,000 that you screen, 3 of them are going to have a major complication. And twice as many will die as a result of screening as will have been saved. So this screening strategy does not work when you take low-risk people. This is true of any screening test. So how do we measure risk? It turns out that there are number of ways that we can identify lung cancer risk. None of them terribly accurate, but over time, investigators have developed a number of models, if you will. That can take clinical characteristics of a patient and predict their long term risk in developing lung cancer, and it used to be just a simple as asking people whether or not they smoked. We now know that screening, or rather that smoking, does not equally distribute the risk for lung cancer amongst people who use tobacco. So some tobacco smokers are going to be at far greater risk than other people who smoke the same amount. Our ability to distinguish amongst them is very poor, but it's getting better. And this is a model published by Martin Tammemagi from Brock University in Canada. It used data from the screening trials to look at who developed cancer, who died from cancer, in relation to their baseline characteristics. Things like age obviously matter, smoking history still matters. But additional factors which we can easily identify upfront, including the presence of emphysema. A personal history of any kind of cancer or family history of cancer. Race and ethnicity have an impact on individual risk as does the duration of time that you've been an ex-smoker, if you've managed to quit. And so all of these can be plugged into an equation and give us data on, for instance, down here you'll see a number that predicts this person's risk over the next 6 years, 1.2% risk of developing lung cancer. So, with this in mind, I want you to think about screening biases. And here's a question about how well we understand the biases that are inherent in each study that uses a cancer screening test. Which of the following are true of screening study biases? A, Lead time is necessary for an effective screening program. B, Lead time bias cannot explain differences in survival. C, Length bias can falsely reflect a mortality benefit in a screening program or D, more early stage disease in a screened group indicates the screening modality is effective. So this is a tricky question. Think about that and we'll come back to that question in a minute. So here is a curve that I like to use to talk about screening biases. In a basic concept of any screening test is that whatever disease you're screening for has some pre-clinical period. That is a period where, perhaps it's detectable with some test but before it can be detected by the presence of symptoms. Now, every disease also has a critical point, that point during which treatment might still be effective. And an example of a disease for which screening can't be affect is one we called mad cow disease. This is a disease that once you have it, if it can be detected, it's almost universally fatal. This critical point occurs long before the disease can be detected by screening, which is why we worry so much about an awful outcome, like mad cow disease. Now, cancer fits somewhere in the middle. Cervical cancer can be detected by PAP smears, long before a woman will present with signs or symptoms of invasive cervical cancer. That critical point occurs at a time when the disease is still treatable, even preventable. Colon cancer, the same way. Breast cancer is a disease we currently screen for, and there's a lot of controversy behind that. But most primary care doctors would never question the utility of breast cancer screening because we've been trained to believe that screening for breast cancer helps woman who are at risk for breast cancer. So cancer screening, the critical point where the disease is still treatable, occurs at a point where it can be detected by screening before it is evident by signs or symptoms of the disease. And finally, for completeness sake, there are many diseases for which screening is unnecessary. Community acquired pneumonia, signs and symptoms of community acquired pneumonia typically occur At a time when we can still intervene in the natural history of the disease with antibiotics. So think about screening in these terms, and it becomes much easier to understand the bias. So a group of investigators wanted to understand how well internists, practicing internists. And in this case, practicing internist attending a national meeting. How well they understood these bias. So they presented these internists as part of the study with two hypothetical screening tests. One, in which no screening was performed. The disease started, cancer was diagnosed based on symptoms at a certain age, and it killed everybody who got the disease. On the other hand, if you have a screening test which can detect that disease earlier, even if screening is not effective, all of those people are still going to die at the same time they would otherwise have died. They will just have known they had cancer for longer. What you have here as a screening test that improves survival but does nothing about the mortality of the disease. This graph depicts why mortality is the only effective measure of the ability of a screening test to intervene in a disease. Survival is a great measure is you have a drug that treats a disease. It's not a good measure of the effectiveness of a screening test, because screening will always falsely prolong the survival from the disease. So you cannot use survival as a means of measuring the effectiveness of a cancer screen test. At the same time, there is a bias called over-diagnosis. And here is a graphic example. In the upper part of this figure, you've got a thousand people with a cancer that's ultimately destined to kill them. Five years later, 600 of those people are dead, 400 are alive. That is a 40%, 5 year survival rate. Imagine now that you have a screening test. But that screening test detects all sorts of cancers that may never have progressed. We call this non-progressive cancers. We also refer to this as over-diagnosis. In this case, those same 600 people are still dead from their progressive cancer but we've got 2400 survivors. So, five year survival has improved because you've detected disease that was never going to otherwise become evident. This is over-diagnosis, another bias that is inherent in any cancer screening program. And there are concrete examples of this, breast cancer, cervical cancer, prostate cancer, colon cancer. And it is just as true of lung cancer screening that there are examples of cancers that look, under the microscope, indistinguishable from the other types of cancer that we worry about that would never have become evident in the lifetime of that patient. So, going back to that question, which of the following are true of screenings study biases? The correct answer is that lead time is necessary for an effective screening program. You'll have to have a time during which the disease is detectable before symptoms reveal it's presence. Lead time bias cannot explain differences in survival, that is absolutely not true. It is the cause of differences in survival. Statement C, length bias can falsely reflect mortality benefit in a screening program, nothing can falsely reflect a mortality benefit. So, a mortality benefit is the only, the only thing that you can find that can justify the use of a given screening test and that cannot be faked if you will by length bias. In more early stage in a screened group indicates the modality is effective, that's not true. That is an indication of both length and over-diagnosis bias. You have to be able to understand these biases if you're going to discuss the risk and benefits of screening with your patient. And these aren't just theoretical. So, let's go back and look at the NCI, chest x-ray screening trials which were conducted in the 1970's. These are huge trials, three different trials. Each of which enrolled nearly 10,000 patients, 30,000 subjects were enrolled in the screening trials and in these trials, chest X-ray detected more cases. There was more early-stage disease. Survival was clearly improved in the screen group, but there was no difference in mortality. And this is now a clear understanding that chest X-ray is not an effective screening test for lung cancer. It will not reduce the chances of your patient dying from lung cancer. You cannot screen for lung cancer with chest X-rays. So, all this talk about lung cancer screenings and understanding why we can employ this in high risk patients. The 800 pound gorilla is that tobacco cessation is the most effective means to the end sought through lung cancer screening, that is if we want to reduce the number of people dying from lung cancer. Nothing will achieve this end more effectively than getting more people to quit smoking. So we're screening as about a 20% effect on lung cancer specific mortality. Effective tobacco cessation could cut that risk in a population by up to 90%. So, when we try to employ tools that reduce people's' chances of dying from lung cancer, let's keep in mind that we have two tools. One is more effective and less expensive, but these are complementary tools. They are not either or. They should be used in conjunction. So, going back to our original test question, which of the following findings should support the use of the described screening test? A, a screening test which improves five year survival of screened patients. When we talked about this does not prove that screening test is effective, and B, a blood test which detects more cases of cancer that it's intended to find. That's prostate specific antigen and we don't use that anymore because it doesn't reduce prostate cancer mortality. It is not an indicator of the effectiveness of that test as a screening test. And finally, a cancer screening test which finds more cases of early stage disease compared with a control group, again, an indicator of length bias and over-diagnosis bias, that can be found even in the absence of mortality benefit. So, the only one, the only finding that supports the use of the described screening test is D, which is in fact the finding that we got from the lung cancer screening trial, a 20% reduction in mortality of the screened disease in the screened group compared to the controlled group. And these numbers trip a lot of people up because they look like small numbers. But when you apply them to millions of patients, that 20% reduction is an enormous impact on overall mortality, which is why we should be employing, properly employing lung cancer screening. And that's all I have to say about that. So, in conclusion, lung cancer screening is beneficial for selected healthy, high risk persons. We shouldn't be screening people who walked in to our office with multiple other comorbid illnesses related to smoking. It's only the healthy smokers that can truly benefit from this. Identifying those at high risk is necessary if we want to do this on a high quality public health level or ability to identify risk is critical. And the risk of cancer screening affects everyone. In other words, no matter how high the risk it is for cancer, the harms of screening are going to apply to it. So, you cannot take low risk people and expect them to benefit form this potentially harmful test. The only people who justify those harms are those who are at high risk for the disease that you're screening for.