>> So it's my pleasure to introduce today's Friday afternoon lecturer, Amy Berrington de González. Dr. Berrington has been at the NIH now for four or five years. Her undergraduate training was in Manchester in England, where she specialized in mathematics, and then she has a Master's in applied mathematics and a Ph.D.(D.Phil.) from Oxford in cancer epidemiology. After that, she spent some time on the faculty at Oxford and then came as a tenure-track investigator to Johns Hopkins, shortly after which we were delighted to have recruited her to the Radiation Epidemiology branch at the NIH, and in short order, she became a tenured senior investigator. Her interest over the last many years has been in understanding the effects of medical radiation -- the health risks of medical radiation -- particularly the effects on cancer. She's been involved in lots of kinds of international studies and, in fact, the Radiation Epidemiology Branch under her leadership has come up with a new way of calculating the medical effects of radiation. And so she's going to talk to us today about medical radiation and cancer risk, assessing the price of progress. Amy. >> Thank you, Dr. Gottesman for the kind introduction and for the invitation to come and speak today. So first the progress, then. Medical radiation is undoubtedly one of the greatest advances in medicine in the 20th century. If we look at the very earliest x-rays of Roentgen's wife's hands in 1897, and then think about the progress through to fluoroscopy, to CT scans, nuclear medicine, interventional radiology, intensity-modulated radiotherapy, amazing cardiac heart mages and proton therapy centers. So there's no doubt that medical radiation saves lives, prevents unnecessary surgery, and detects disease early. It's also established, though, and it only took a few years to become evident that there are downsides to radiation exposure, in particular, the risk of inducing a cancer. This is, as Dr. Gottesman said, what I've spent the last 15 years of my research career primarily focused on -- understanding how medical radiation fits into this picture of the major causes of cancer in countries like the U.K. and the U.S. today. Now I'm going to talk about the research that I've done in both the area of diagnostic exposures and also therapeutic exposures. So as I said, it was very early on that we realized that actually these amazing new x-rays also probably had downsides. Already by 1902, there were skin cancers being reported on the hands of radiologists, and it seemed very likely these were a result of the x-rays because they tended to put their hands into the beam every day in order to test their equipment. By about 30 years later, there was some case series of leukemias reported in radiologists and as a result, two major cohorts were set up that really were the first studies to confirm these risks. And it was actually also working on studies of radiologists that was my first introduction to radiation epidemiology. When I joined the cancer epidemiology unit in Oxford in 1998 to do my Ph.D., I was sent off on my first day to Richard Doll's office -- a fairly well-known cancer epidemiologist -- to talk about doing a follow-up, an extended follow-up of his cohort of British radiologists, and that was my first doctoral thesis project and resulted in the publication shown here. It was actually reading another of Dr. Richard Doll's classic papers that gave me the ideas for the next project for my doctoral thesis. So his 1981 paper, the 200-page paper on the avoidable causes of cancer in the U.S. Tucked away in there on one of the pages is one line that says something like "and medical radiation is probably responsible for about half a percent of cancer deaths." And I'm reading this, I was pretty surprised actually, that seemed quite high to me, and I went to talk to Richard about possibly doing some follow-up work on this observation that he had, this estimate that he had, and updating it. Of course since 1981, there had been a lot of changes in diagnostic radiation, use of CT scans had only just been introduced at that time. So I was interested in what the impact of those changes would be. And he was very enthusiastic about this idea, especially when I pressed him a bit on the methods that he'd used for that calculation, and he admitted that it was probably a back of the envelope calculation and he didn't really have the details any longer, and so he thought it was a very good idea if someone did something a little bit more systematic. I have to say, though, although he was enthusiastic, I think most of us in my department thought I was a bit crazy. They were working on sexy things like diet and cancer and HRT and cancer, and they thought radiation was a bit passé. We already knew it caused cancer, what more did you need to do. I think the following 15 years eventually proved that my decision wasn't quite so crazy as they thought and it resulted in a very interesting career. So a little bit of background on certainly what we knew about ionizing radiation and cancer at that time when I was starting my work. Of course most of what we knew comes and still to a certain extent did come from the atomic bomb survivors, the long-term follow-up of this cohort. So of course it's an established carcinogen, and certainly at least within the range of doses experienced by the atomic bomb survivors, the risk is clearly linear with dose as shown in the figure here. The risk remains elevated throughout lifetime, and we've seen that it can cause probably most types of cancer. An important observation is that the risks tend to be higher for the younger age at exposure. And there are interestingly some organs that seem to be more radiogenic. Still, there are a lot of uncertainties, and some of the key uncertainties are in the dose range precisely where most of the medical exposures occur, which is in this low dose range below .1 Gy and the high dose above about 5Gy. There's uncertainties about what the shape of the dose-response might be in those ranges, and even if, indeed, there might be any risks. There have been, though, a number of studies already looking at medical radiation exposures, even at that time. In general, they tend to focus on special populations that had high levels of exposure, so although they had diagnostic exposures in each individual exposure had a small dose, for example, the tuberculosis patient had some, even hundreds of fluoroscopies, so the cumulative doses were quite high. But these studies did establish at least the general principle that these types of exposures, diagnostic x-rays, can indeed increase your cancer risk. It's difficult, though, to study exposures in the general population the sort of more typical exposures that most people receive, which are just a few x-rays, perhaps throughout their lifetime. To study those sorts of exposures directly, given that the individual risks are small, would require extremely large populations followed up for very long periods of time. So mostly what we have done then to try to estimate what the risks might be is to use risk modeling techniques, which are based on the Japanese atomic bomb survivors and also some of the studies shown here, and it's what we call risk projection. So that was what I used then for my doctoral thesis work, to estimate what proportion of cancers might be attributable to diagnostic x-ray use, and this was levels of use in the early 1990s, when I was doing this work -- this was the best available data that we had. So this figure shows the results from four of the 15 countries that I studied and effectively shows the range of estimates. So the U.K. had some of the lowest levels of use and, therefore, the lowest estimated risk, just above half a percent, so still actually close to the estimates in the U.S. from Richard Doll's paper in the early 1980s. And Japan already at that time was up more than 3%. The U.S. was somewhere in the middle, around about 1%. So at this time, Japan was already using a very large number of CT scans, and that's the reason -- one of the reasons that its estimated risk is so much higher. This should have, I think, given us a warning about what might happen in other countries, but for some reason, we just seemed to think at that time that Japan was sort of an outlier and everyone else was sort of in the normal range. What we didn't know, though, because there hadn't been any data published for a long time, there aren't national surveys conducted in the U.S., was that although in 1980, there were 3 million CT scans, but by 2007, there were 70 million CT scans performed. And there was no data effectively in between those two data points. So for 25 years, we didn't really know what was happening. Although now we've gone back and reconstructed the history, we see this is what was happening -- that from 1980's onwards, there was a gradual upward increase in the number of CT scans, and after 2000, there was really dramatic rise, more than 10% increase per year, and this basically coincides with the introduction of multi-detector CT scanners, which could perform the procedures really quickly. So as I said, this dramatic change didn't become apparent until this report was published in 2007, but as soon as it was published, of course, then the question was, what would the impact of these changes be on those risk -- those previous risk estimates we'd conducted for the U.S. It's not just volume of procedures that was a concern with CT scans, it's also the dose levels. So we're still in the low dose range, but the dose from a conventional x-ray to the same parts of the body, the CT dose is typically on order of magnitude 10 times higher or even potentially more. So the combination of the greatly increased volume of use and the fact that these procedures are higher doses than conventional x-rays meant that we were certainly wanting to assess what the impact was of these changes. So with the team that conducted that survey that I described, team of radiologists and health economists, and statisticians, we looked at what the estimated number of future cancers might be from these 70 million CT scans, and again, we used these risk projection methods, excluding the proportion of CT scans that we estimate were taken just in the last few years of someone's life, and also the CT scans that were actually taken in relation to a diagnosis of cancer. That turns out to be only about 15% in total, so it's not as large a contribution as many people think. So we excluded those, and still estimated that there could be an additional nearly 30,000 future cancers from one year's worth of CT scans. The figure here shows the breakdown according to the type of CT scans. Abdomen and pelvis are by far the most common in the U.S., so as expected, they make the largest contributions. What this report also highlighted, this survey, was actually that all other types of medical diagnostic imaging had increased as well. Although CT had increased the most, that conventional x-rays had actually also increased and nuclear medicine and interventional radiology. A particular component of nuclear medicine that we were also concerned about is the use of cardiac stress tests. So they increased nine fold between those two surveys, and actually their doses can be even higher-- certainly double if not up to five times higher than -- say an equivalent chest CT scan. Despite the fact that these are typically performed in older individuals, if again we use the risk modeling techniques, we estimate again an additional seven and a half thousand cancers just from one year's worth of cardiac stress tests. So an area where there have already been concerns raised much earlier, back in 2001, was the use of pediatric CT scans. There, the issue wasn't so much about the volume of use, but making physicians aware that because children we know have higher cancer risks, and also they tend to have higher radiation doses in CT scans. And that was particularly true back in the earlier periods of CT use, because it was very common that the machines weren't adjusted for the size of the patient. Even now, with adjustments, their doses tend to be somewhat higher. So this was an area where it was actually going to be potentially feasible to study the risks and also from a public health perspective, important to do so. So people were asking really is there a risk from these procedures and if so, how big is it? Can we again predict it based on the atomic bomb survivors or should we try an study it directly? Given the higher radiation doses, higher radiation risks and lower background levels of pediatric cancers, you actually have potentially the statistical power to do a study to look at these risks directly, which is very difficult to do in the adult setting or with conventional x-ray procedures. So the collaboration was established between the U.K. and the University of New Castle in the U.K. and NCI to actually set up a pediatric CT scan study. And the system in the U.K. both at the National Health System and the National Cancer Registry System meant that it was more feasible to do it there than here, although in some ways, there were more concerns here. But what we did was we used the radiology systems information databases at more than 100 hospitals and then linked these to the U.K. cancer registries to look at the subsequent cancer risks in these children. So from all of these radiology databases, we extracted data on all the patients that had CT scans before the age of 22, which gave us a total cohort of nearly 200,000 children, and by linking them to the cancer registries, we ensured that we excluded all the children who actually already had cancer before they had the CT scan, so this is 180,000 cancer-free patients. We also then actually excluded anyone who developed a cancer or died within the first two or five years of their first CT scan. This varies depending on the type of cancer we're looking at. But effectively we're trying to ensure that the CT scan wasn't taken to diagnose the cancer, it wasn't taken because they already had cancer, and that you've got a real chance that the subsequent cancer has actually been caused by the CT. So we published the first results from this study last summer. We've got now an average of 10 years follow-up on these children, and during this time period, we've got 74 leukemias developed and 135 brain tumors. Those are the cancers that I'm going to show you the results from today. These are the most common cancers so far in our cohort, or the ones that we've got the power to analyze, and also given that they're some of the most highly radio-sensitive tumors, they were the ones that a priori we were most interested in. Out of the 300,000 CT scans in the analysis as expected in this pediatric population, the large majority are head CT scans, but you have got a range of many other different types of exposures as well. So these are the first results. This is the relationship between subsequent leukemia risk and the cumulative radiation dose to the red bone marrow from the CT scans. You saw a clear linear dose-response relationship, statistically significant, and for a cumulative dose of about 50 mGy, approximately a tripling in the risk of leukemia. And very similar results for brain tumors and cumulative dose to the brain. So again, very clear linear dose-response relationship, highly statistically significant, and actually somewhat similar relative risk, around about 50 mGy, and relative risk of approximately 3. The cumulative doses to the brain can be much higher in total because the dose to the brain is much -- from a head CT scan -- is much higher than the dose to red bone marrow. Importantly, these results are also pretty consistent with the atomic bomb data, so if we take the Lifespan Study of the atomic bomb survivors, and restrict those people who were exposed in childhood and also restrict the follow-up period to make it as similar as possible to our study, shown in this table here is effectively the dose-response, the risk per mGy. You can see for leukemia, it was actually very, very similar, and certainly statistically compatible. For brain tumors, it is statistically compatible, but our risk estimate is about four times higher than the Lifespan Study. So we think that this could be two things or a ombination of two things. It could suggest some inaccuracies in our dose estimates and we're doing more work now to try and extract individual CT scans and do some more accurate dosimetry. It could also be an indication that we do still have some bias. Despite our long exclusion period, we still have some children in the study who effectively had their CT scan in the -- a long time before but in the diagnostic workup period to the brain tumor. That's much more likely for brain tumors than for leukemia, because CT scans are typically taken for symptoms of brain tumors, they're not typically for leukemia. So of course most of us can't think in mGy and want the results in terms of number of head CT scans -- or number of CT scans. So if we translate these relative risks into number of CT scans, then this tripling in risk for leukemia for about 50 mgy -- to get to that dose using current CT scan techniques and current settings, it depends on age, but for the infant, about five head CTs will give you that dose, and for older children, about 10. For brain tumors, again for this tripling in risk, we saw that at about 60 mGy, and the doses are less dependent on age to the brain. So typically only two to three head CT scans will get you into that dose category. Something that was really important that we emphasized in the paper, hough, is despite these large relative risks, of course we're talking about rare cancers. These are childhood cancers and they're rare, and the absolute excess risk, therefore, is still small. So during the 10-year follow-up period, we estimated there was about one excess cancer per 10,000 head CT scans, either one excess leukemia or one excess brain tumor. So we think that this study is the first direct evidence of a cancer risk after pediatric CT scans, and the reason we say the relationship is likely to be causal -- we're talking about an established carcinogen, we're not talking about potatoes or ice cream or something, this is something where there's lots of existing evidence of likely risks, clear evidence of a dose-response relationship, and also the results are reasonably consistent with the existing evidence. Also, from a statistical perspective to get to relative risk of more than 3, it's unlikely to be due entirely to confounding. You have to have a very strong confounder to actually create a risk of that size. Nevertheless, we think that the brain cancer risks may be overestimated, and it's certainly important, of course, to replicate our findings and also potentially to try and look at the risks in adults. So fortunately all these replication studies are already underway, and actually in the next year, we expect two of them to also report their first results. So there's a large cohort in Canada and one in Australia that are very advanced and we understand are already also in the analysis process. A smaller cohort in Israel that actually won't be published necessarily on its own ever, but was developed as part of this collaborative project with an aim of pooling all these cohorts eventually. So [garbled] all four cohorts were designed using similar protocols to facilitate pooling, and actually we're holding the first collaborators'meeting this spring here at NCI. So we'll be very interested to see first what the other cohorts are showing, and then if we can put them all together, it will enable us to look much more quickly -- certainly at some of the cancers that are rarer in this population like breast cancer and thyroid cancer. Given the age of the population, there still aren't very many cases, but they're also highly radio-sensitive cancers and we want to be able to try and look at these risks. By putting all four cohorts together, we should have data on more than a million children and we should be able to not only look at the consistency of the effects across studies, look at factors like effect modification by age, and as I say, look at these other key cancers earlier on than any of the individual cohorts can. The Canadians also have a cohort of adults, about 600,000 adults, that's not as advanced but certainly should give us the possibility of looking at these exposures in the future. Even before we published the findings from our pediatric CT study, actually there had been a number of reports by key institutions and international organizations already calling for changes, and raising questions about the levels of use in campaigns like the U.S. campaign "Image Gently" for pediatric CT, and the World Health Organization global initiative on radiation safety. And they all have the same key underlying message, which is important to always reiterate to the patients that in general, the risks from individual procedures, as I've shown you, are small. And so as long as the test is clinically justifiable, then the benefits should indeed outweigh, far outweigh, the risks. Nevertheless, if the procedures are unnecessary, or done with unnecessarily high doses, then there will be unnecessary risks. So they also come up with -- in general -- common approaches to reduce these risks. So the American College of Radiology, for example, has developed appropriateness criteria that are available on their website. You can look up any diagnostic category and they rank all the different imaging procedures according to which is the most appropriate for that diagnosis. Often CT doesn't necessarily come out top, and sometimes the alternatives like MRI or ultrasound have -- are more highly recommended and of course these don't involve ionizing radiation. Another key message from all of these reports is to avoid unnecessary repeat studies, so people being moved from one hospital to another undergoing exactly the same CT scan during the same day rather than just taking the film with them. They also all emphasize the need to try to reduce doses, particularly by standardizing protocols, efforts to monitor doses like that that's been introduced here in the clinical Center, and now there's also interest which, of course, the manufacturers are particularly keen on, in developing new technology to further reduce the doses. Out of all of these, whilst I think that it's important, of course, to reduce the doses, I have to say I'm more concerned personally about the levels of overuse. And the reasons for this or at least some indirect evidence of this is if you look at the current rates of use across the world, and you look as shown here, my four example countries, even though rates of use have increased as well in the U.K., they're still seven times lower than the rates in the U.S.and even in Germany, which is actually the country with the highest level of CT use in Europe, also one of the few countries where radiologists are paid per procedure in Europe, the rates are half what they are in the U.S. Japan still way ahead of everyone. You look at nuclear medicine and there's also a very wide variation in use but interestingly, not exactly the same patterns as for CT scans, which I think just shows you also how there's this regional practice which is developed, and then tends to be followed. So nearly all of this variation in use in nuclear medicine around the world is due to cardiac stress tests. Basically all of the other tests -- the thyroid procedures, bone scans, are similar sorts of rates, and this is all cardiac stress tests. And for some reason the Japanese aren't so excited about those. So what happens, then, to the figure that I showed you at the beginning if we don't change practice and if we don't reduce use, particularly in the U.S.? Well, if we take these updated estimates of use and put them into these calculations, then this isn't what the attributable risks are now, this is what the attributable risks will be eventually if we keep using CT scans and nuclear medicine tests at the current rates into the future. So in some ways this is good news, there is a chance to change this before it happens. So these are the changes in these three countries, and this is the change in the U.S. So if we keep using at current levels, then eventually, we estimate that about 3% of cancers in the U.S. could be due to diagnostic radiation exposures. So where does that put us in this picture, then? Well, currently we're probably already around 1% if not higher, and so that's certainly already in the top 10 causes of cancer in the U.S., but it could go into the top five. So for the last 10 minutes of the talk, I'm going to switch actually to my work on therapeutic radiation exposures. One of the reasons that I was particularly keen to move to NCI, was to work on this area of medical radiation. The Radiation Epidemiology Branch has really been at the forefront of this research for the last 30 years and continues to be so. So as a result of improvements in cancer treatment, the number of cancer survivors is increasing in the U.S., and there is increasing interest and attention now on their long-term health and thinking about ways to also potentially reduce their risk of getting another cancer. So we know from analyses of the SEER cancer registries that cancer survivors in general have a higher risk of developing a second cancer compared to the general population. Most likely a result of a combination of the lifestyle and genetic factors that were probably responsible for their first cancer, but also a result of the treatment for their first cancer. And it's this fact that I'm particularly interested in studying. So one of the first projects that I undertook when I moved to NCI was to try to estimate what proportion of these second cancers then might be related to the radiotherapy treatment. Again, some of my colleagues actually here at NCI thought I was a bit crazy on this occasion -- not so much because they thought that it wasn't interesting, but they thought it was really, really hard. But I think we persevered and many, many thousands of calculations later, we used the SEER cancer registries, which have long-term follow-up and have information, at least basic information, on treatment to compare the patients who had radiotherapy with those who didn't have radiotherapy. We looked at about 15 types of first solid cancer that's typically treated with radiotherapy in adulthood. This figure here shows the overall the percentage of second cancers we think that might be related to radiotherapy which overall in the population was just under 10%. It also shows how the proportion probably varies according to the type of first cancer, being lower for cancers like oral and breast cancer, and higher for prostate, cervix and testes. This is most likely because these cancers are all -- receive pelvic radiotherapy, where you've got many more organs in and close to the radiotherapy fields. With cervix and testes, it's also because the patients are on average younger adults than for most of the other cancers, and as I've already said, we know that younger age of exposure gives you higher risks. Again we have to think about the absolute risks, not just the relative risks or the attributable risks. And overall we estimated that by about 15 years after the treatment, there were about five excess cancers per thousand patients. So this is the sort of estimate that can be used to try and put the risks and the benefits of the treatment into context for the patients. If we think about breast cancer, for example, we know well from the radiotherapy trials of breast cancer that the benefits from radiotherapy are about five breast cancer deaths prevented for every hundred women treated, so here we can see, then, that the risks are on order of magnitude 10 times smaller than the estimated benefits. I still think, though, and this is particularly for cancers like breast cancer and prostate cancer, where we're increasingly concerned about overdiagnosis of these cancers due to screening that that calculation is not always going to be so straightforward. So for these cancers, if we think there were high levels of overdiagnosis and that those patients actually didn't really need to ever be treated, that they would have survived for the rest of their lifetime without treatment, then again we're in a situation of potentially all risk and no benefit. By definition, it's difficult to study the late effects of new technologies quickly, and so one of the drawbacks of that study was that we were using data primarily from patients treated in the 80s and 90s, and as with diagnostic radiation in the last decade, there's been a revolution in radiotherapy as well. So this graph here shows the change in the use of conventional radiotherapy over to intensity modulated therapy for prostate cancer patients. In only four years, there was a total reversal from the majority receiving conventional to the majority receiving intensity modulated radiotherapy, which is a much more targeted technique in theory. Again, very typical of the U.S. to introduce these technologies extraordinary rapidly. Another key change which is really happening in particular as we speak is the introduction of proton therapy. Nothing like at the levels yet of IMRT, it's much more expensive and requires enormous new facilities, but between 1960 and 2004, there were just two centers operating in the U.S. In the last seven years, there have been nine new centers opened and as I say currently as we speak, there are nine new centers around the U.S. under development. So these new technologies certainly in theory should have great additional benefits. The theory of them, they work in different ways but the general idea is that they're much more targeted so there should be more effective at treating the cancer, but also they should reduce the amount of normal tissue that gets high-dose radiation exposure, so you should have a reduction in the number of acute toxicities, although there is -- that's the theory, it's still not very clear yet from the evidence that -- the direct evidence -- that that's what happens, but that's certainly a good theory. From the perspective of the second cancer risk, it's not very straightforward to assess what the balance of these changes in the doses will be. So you should reduce the amount of tissue that receives high-dose exposure, but both technologies, for different reasons, actually increase the amount of body that receives an increase of the amount of low dose exposure to the rest of the body. And with proton therapy, one of the particular concerns, it depends how it's delivered, but it can actually give a whole-body neutron dose. Neutrons, we don't understand them very well, but there's certainly lots of experimental evidence that suggests that the cancer risks are much higher per unit dose of neutrons, possibly 10 times higher. So whether these new technologies will reduce or increase second cancer risk is currently unclear. There's been a number of modeling studies that have tried to predict what might happen and they come up with conflicting results. Some suggest they will increase second cancer risks and some suggest they could decrease them. And the jury is currently out. But we're certainly working very hard to develop patient cohorts, we think this is an important question to study. We're looking at databases like SEER-Medicare. Some of the state cancer registries have started collecting data on type of radiotherapy delivered, so we're hoping that we will soon have a sufficient and good database in order to look at these questions. And we're also trying to update some of the modeling techniques so that we don't have to wait necessarily 10 years to answer this question, but that we can get some good insights in advance as to whether these really are safe enough. So in conclusion, I think that I showed that my colleagues in Oxford 15 years ago were wrong. I wasn't crazy to spend the next 15 years of my life studying medical radiation exposures. And that's one of the key reasons, is it's probably in the top 10 causes of cancer in countries like the U.S., and that's both in the general population and in cancer survivors, the therapeutic effects probably put radiotherapy even in the top five causes of second cancers in those patients. So it's really important that we have information on the potential risks and can quantify the risks in order to assess the balance of the risks and the benefits. Particularly in the U.S., the technology changes so rapidly, and use can increase so rapidly that it's important that we conduct studies of these new therapies for both clinical reasons and because of public health concerns. And that's really what the Radiation Epidemiology Branch is -- one of its key missions. So finally, actually it's a really important tool as well for understanding radiation carcinogenesis more generally. As I showed you at the beginning, most of what we know comes from the atomic bomb survivors, which is within a very specific dose range, and it's really medical radiation studies that have given us -- that in the future will certainly give us more information about the very low-dose exposures and also the very high-dose exposures. It also will potentially give us information about different types of radiation, and because the exposure history is often contained in medical records, this can be a much more accurate assessment of radiation dose than, say, we can get in some of these environmental studies, where you're trying to model what happened in fallout from nuclear weapons tests. So I'd like to thank the very large number of collaborators that were involved in the studies that I have discussed today, and finish with this list of the key references, and thank you for your attention. [applause] >> Questions? And I'll ask, if you have questions, to please go to the microphone since we are broadcasting this event. >> A simple question. What is the definition of brain tumor? Are you including meningiomas as the current WHO definition includes meningiomas as a brain tumor, which is a benign tumor, and so in considering the risks, if we're talking about the glioma series, which has been shown to be radiation-induced in animal models. But also meningiomas are very well known as being a second tumor from radiation therapy, so if you could clarify that? >> Yes, so it's out of those 135 tumors, I think that the major subgroup is gliomas, 65 gliomas, but we also did have, we did include meningiomas and schwannomas and there's about 20 of those or something in there at this point in time. >> Okay, thank you. >> Excellent talk, Amy. Thank you. You estimated that combined risk from one head CT was approximately one in 10,000 of leukemia and brain tumors. That was over a 10-year period. Would you care to speculate about lifetime risk from one head CT? >> Yes, so that's a good question. As I said in the beginning, the risk of radiation-related cancer actually remains elevated throughout the rest of your life. So it's true, the total risk, lifetime risk is going to be higher. If we use the risk projection models, I think that it comes out at something like one per thousand head CTs, rather than one per 10,000, so eventually 10 times higher. But still a small risk, compared to the benefits. >> And of that one in a thousand, what percent is leukemia and what percent is brain tumors? >> I think -- well, for that actually, that one in a thousand is all cancer types, that's if we do the extrapolation to all cancer types. >> All right. >> I think leukemias and brain tumors are a relatively small fraction when you go out across the rest of the lifetime. >> Thank you. >> Thank you for the presentation. I'm a health physicist, so I've seen lots of this before, and I'm not sure if it's a general comment, but very often when we look at these numbers, we look at numbers of CTs. The problem that I have with that is that we know effects are due to dose. Has any effort been made to quantify the doses received as opposed to just the number of CTs? The best example I would consider, up until about maybe the early 2000s, all children were receiving brain scans using adult techniques. Which are generally 3 to 4 times higher than the current numbers we're getting now. So my concern is that -- is there a bias because we're looking at people who have received CTs in the past who received higher doses in the past, and how can we -- like I say, is there any consensus of trying to do something of this nature to try and quantify these numbers better? >> So the results that I showed you from the pediatric CT scan study were in terms of dose, not in terms of number of CT scans. I then translated the risks into number of CT scans because that's what most people can deal with better. So those -- the figures I showed with the dose-response relationships, those are based on estimated organ doses for the CT scans according to age, the calendar period of the CT scan and the type of the CT scan. So as much as we could, we tried to reconstruct doses, taking into account the issues that you are raising. Then, when I convert the risks back to number of CT scans, that was using current settings and current dose estimates. So 60 mGy from current settings, 2 to 3 CT scans in a child to the head should give you about 60 mGy. >> Thank you. >> That was an excellent talk. Thank you very much. One question I have, perhaps at the NCI level, is if you could eliminate 30,000 cancers a year, that would be considered a huge achievement. So what direction is the Cancer Institute looking for trying to substitute ionizing radiation in diagnostic studies with alternative approaches? >> So I think that's a very good question. Unfortunately, probably we can't avoid all of those 30,000 cancers, as you're pointing out. But I think that reducing use, primarily, really could reduce some of those unnecessary cancers. I think that seems to be an issue that's very difficult. I'm not sure what the Cancer Institute is doing, particularly. I've been to lots of conferences and workshops over the last year with organizations like the American Board of Radiology. They're talking about these issues and trying to come up with suggestions, but it's a difficult issue in this country, particularly when you start talking about not doing procedures, that rings alarm bells of denying people care. >> Right, but perhaps that's where the National Institutes of Health really should step forward. I've been impressed. Your data is a very good example that you can see the problem most easily in the pediatric populations, but the same problem applies to older patients as well, just at a slightly lower rate of cancer incidence. We have a collaboration with Children's National Medical Center, and as part of that, I saw that their development plan for radiologic imaging uses MRI at a 4:1 ratio relative to CT scanners. Here at the NIH, the ratio is closer to 50/50 in terms of scanners, and when you think about the throughput of patients going through those different technologies, it's probably closer to 2:1 or 3:1 patients getting CT scans rather than MRI scans. So I think if we continue to learn from our pediatric colleagues and analyses like what you're saying or what you're presenting, you know, we really -- I think there are tremendous opportunities for replacing diagnostic tests that use ionizing radiation with alternatives. >> Yes, I think you're right. And actually, I gave a talk recently in Europe and some of the radiologists came up to me afterwards and said we're already trying to avoid doing head CTs in the children in the emergency rooms. We keep them in overnight instead, for observation, and then by the morning, the MRI scanner is up and running and the MRI scanner person is in, so if we still think they really need imaging, then they go into the MRI scanner. But my understanding, is that that's going to take some time in the U.S., to make those sorts of changes. >> Amy, could I help answer his question, too? >> Sure. >> What we do here at NIH, I don't know if he's listening -- but so I'm the radiologist lead for CT here at the Clinical Center, and we're not working together yet with NCI, but what we've done is we've taken the Joint Commission Sentinel Event Alert Number 47 that came out in August 2011 and we've taken each of those issues. And if you look at the first issue, in JCSEA 47, it's "Right Test", and it's right along with the slide you showed, with the Choose wisely, Image wisely, and the first thing is "consider alternative tests". So when I protocoled about 50 CTs today, before they were done, a radiologist protocols them to make sure that they're indicated, and we're in touch with Children's and everybody else too, on how we are lowering dose but also preventing exams that can be done by these alternative methods, and it is working -- is to look at that first. So we make another judgment call in addition to the ordering exam. So that's what radiologists are doing, thanks to studies that Amy has done here. It's admirable work -- thank you very much. So if that helps answer your question. And with kids, too, we've made a concerted effort talking to Hopkins, Children's, Yale, everybody, we're all in groups together with the AACR, where we've cut doses in kids, for example, down 50% from where we were two years ago, and that was just by lowering kv in these kids that were getting repeat CTs. They still needed chest x-ray, so that wasn't a good enough alternative, but we gave them doses that were half the amount, and no CT has been repeated. So that's what we've done here at the Clinical Center. Thanks again for that talk. >> Thank you. >> Hi - sort of two entirely different questions. First one being I did my postdoc in France and I had the fortune to get hurt several times while I was a postdoc in France. And one thing that I noticed with the integrated electronic medical records, that they kept a tabulation of the acute radiation dose and then the longitudinal radiation dose, and so I was wondering, to further complicate the analysis of your data as to looking at the differential timeframes of acute dose, longitudinal dose, longitudinal dose within particular time periods and then also at different regions of the body and of how -- because I would assume that there's a difference of one's cancer risk, if I got 10 different standard x-rays all around my body than 1 CT in one space, of how that gets integrated into your analysis. >> So it gets integrated again through the organ dose calculation. So that's precisely how we take into account if someone has a head CT and a chest CT and an extremity CT, then we estimate the red bone marrow dose for each of those tests and then cumulate them. If they all happened on the same day, they will go into the risk calculation, the analysis, as being a single dose on that day. If they happened over five years, then effectively the dose only gets taken into account at the moment that it occurs, so the longitudinal nature of the dose and the differential types of exposures both get taken into account in the analysis. >> So one thing I was wondering, are you aware of any of the policy decisions that happened in France that enacted some kind of record-keeping like that? I mean, it seems like policy was leading the way in that area, and it's lagging here in other areas, where it's not so integrated. >> Yeah, I didn't know actually that they were already monitoring doses like that in France. So it's interesting to hear they can do it, because there's a lot of discussion about doing it here. I was just on a National Academy of Science panel last year where we were discussing exactly this issue and the conclusion and our report in the end says it's unfortunately just not feasible at this time. >> Certainly sounds like it could be a right dataset to play with. >> Yeah, thank you. >> I was wondering if there is any way to get a handle on effects of essentially secondhand exposures to nuclear medicine tests in particular. So for cardiac stress tests, obviously the benefits are obvious, but those people have families, and so now you're exposing, for example, young children to radiation that has no benefit to them, and is -- I just don't know what the level of exposure is. On one hand it's obviously not in that person's body, but we're talking about young children, where you have a much higher sensitivity. Is there any way to get a handle on that kind of issue? >> That's an interesting question. For the cardiac stress test, at least my understanding, is that would be really minimal. So I think that there is more concern, say, about the thyroid treatments, which is one of the reasons that they suggest keeping some of the patients in hospital and there have been questions about the exposure, certainly to the nurses and other people who tend to them, but I think any estimates that have been done of that still suggests it's really a very small dose. >> Thanks. >> Amy, one of the issues in the field of ionizing radiation has been whether there is a threshold lower dose below which normal repair mechanisms can compensate and so there's no net biological effect. So it seems to me that your data might allow you to look at that. For example, for minimal radiation dose, let's say somebody gets a chest x-ray every five years or something like that, and no other procedures. Is there any evidence in those populations that there's an increased incidence of cancer to the target organs, let's say the chest or something like that? Do you have enough data to answer questions like that? >> So no, not yet. So certainly at the moment, the power in those lowest dose categories wouldn't allow you to do that, but on the other hand, you look at those lower dose categories and you see that although the confidence intervals are very wide, the point lies exactly on our linear curve. >> There's no safe lower dose? >> It certainly doesn't -- it doesn't suggest it, it doesn't -- we can't confirm that there's no threshold base in our data but they certainly don't suggest it. Once you put all those studies together and have a million children and have some longer follow-up, we might be able to get more down into that dose region that you're all talking about. >> So one thing we used to do in subjects who were here for experimental procedures or on protocols, was we would try to compare the radiation dose to the dosage that would happen in regular life. Obviously there's some radiation, cosmic radiation on earth, and as you go higher up, the radiation gets higher. So we were just talking about this before. People who are involved in lots of transcontinental flights, for example, or even flights across this continent, are exposed to significant amounts of gamma radiation, right? Is that -- can your data -- do you have enough power in your analysis to say that somebody who flies to Europe five times a year is going to have an increased incidence of cancer? >> Again, not directly at this point, but the fact that risk estimates were so consistent with the atomic bomb survivors, that gives us extra confidence, I suppose, that you could use those existing risk models to do those sorts of projections. >> And how does such a flight compare to a chest x-ray or a CT or what are we talking about in terms of magnitude? >> I think it's -- I can't remember offhand, my impression is that most of these flights, whilst they're not no-dose, they're more like x-rays, they're nowhere near CT scans, but -- so another point on your question, though, about the background radiation question, actually this same -- this last year, just after our CT scan study was published, there was the first study was published in the U.K. which actually also showed pretty good direct evidence of increased leukemia risks from background radiation. This was a large case control study of more than 5,000 children in the U.K., where they went back and estimated, depending on where they lived, what their background radiation was and their total risks. >> I remember a study a while ago about potassium 65 which people have in their bones, and that if you slept close to somebody, all of your life, your total exposure was additionally great enough to make a difference in cancer risk. >> They actually did see this in the study. Again, the leukemia dose-response was very similar to the atomic bomb survivors and very similar to our CT study. >> Okay. >> Thank you very much for your talk. I was wondering if as a epidemiologist who's clearly an expert on this topic, you'd like to comment on the backscatter scanners being used at our airports now? >> Yes. I do get asked about those quite often [background laughter] so I have done some calculations, and if I can remember it offhand, the doses are really very, very small, and I think they are 10,000th of a chest x-ray, if we understand the dosimetry properly. There was some question about whether the traditional dosimetry really applies, but if that's correct, then you have to go through it 10,000 times to even get up to the same dose as a chest x-ray, so it's really very, very small. >> Do you go through it? >> I do, yes. I went through twice at Christmas. >> Can you comment just briefly on the risk of dental x-rays? >> So they're also pretty small, not as small as the airport x-ray scanners. And if you think about the region that's typically exposed, there's not many organs that are going to be really in the beam. There have been a couple of studies looking at thyroid cancer and salivary gland tumors, and they did suggest that based on exposures really a long time in the past, in the 30s and 40s, when doses were much higher and they were doing lots of these panoramic x-rays, that you might get increased -- small increased risks of those tumors, but really based on current techniques and current practice, there should be really, really minimal risk. >>[inaudible] >> Yes, there's also been some studies which suggested some small increased risk of meningiomas. And the other thing about most of these studies is they're case control studies where they're asking the patients to recall their past history of these procedures, and so that raises some concerns about the validity, but we have in our branch a large cohort of rad techs that we've been following up for a long period, and so we have some prospective data, not only on their occupational exposures but also on their reported history, and in the next few years, we hope to look at exactly that question, but in a prospective nature to see if it really holds up. >> Okay. Well, Amy, thank you for stimulating a great discussion, and I think a lot of people here are going to go out and think about their daily practices. Thank you very much. [applause]