 Okay, so thank you all for coming. I think we will, we have a nice room in terms of sound, but we'll project as much as possible. So it's my distinct pleasure to introduce to us here at UCI, Dr. Veronica Vieira. She's an associate professor in the Department of Environmental Health at Boston University School of Public Health. She has a very, very interesting background for those of us interested in interdisciplinary work. She is an environmental epidemiologist, so for those who don't know, you can combine these two fields. It's quite possible, and in fact, Dr. Dean Becker also talks about that as his own field. And there is an international society for environmental epidemiology, and Dr. Vieira is in fact the secretary-treasurer of that society. It's the lightest organization in the field, represented from 15 countries or so. Dr. Dean Becker is also the president for this period. But Veronica got her masters in science in environmental engineering from Stanford University, and then went to get a doctor of science in environmental health at Boston University School of Public Health. As she was just mentioning extensive knowledge of geographical information science in groundwater modeling, cluster detection methods, and pollutants that are persistent in water systems. Some of us in the audience are very interested in these kinds of pollutants. I know one of my students, Daniel, is very interested in the sources of pollutants in water, and I don't know who is here, very interested in PFOA and some of these compounds that we've talked about. Dean, we just talked about you, so good timing. Anyway, so it's really a great pleasure to have our visitors all the way from where it really snows, and I hope you are digging out from one of the biogas this time. She's going to talk to us today about breast cancer risk. There are many, many faculty members who are interested in cancer on campus. In fact, we're really organizing to build that even more so in population sciences. But in this particular case, it's a risk factors from environmental exposure to drinking water that's contaminated. So thank you very much, and we look forward to sharing. Maybe before 12, we will discuss and then we'll go to lunch to continue. Thank you all for coming, and thank you for that introduction. As Deli alluded to, I'll be sharing with you some of my research that I've done on breast cancer risk. And this is really a story that started about 10 years ago when I started as a doctoral student with some of my research there, and it's just sort of progressed through postdoc and now faculty. And again, as I explained to somebody before that showed up earlier, it really is a progression where each study kind of leads to more questions, and that kind of leads to the next step in your research career. And so I'll be sort of taking you through my own story of how that's played out. Feel free to interrupt if you do have any questions that really need to be answered immediately. Otherwise, I think we'll have some time at the end to have some questions and answers. And again, this is research that I've done related to the Boston University Superfund Research Program. So for the overview, there are really three main points that I'll be discussing today. The story begins with a study looking at tetrachloroethylene contaminated drinking water and how that related to breast cancer risk. Using that data set that we collected to address that question, we then moved on to looking at space time analyses of breast cancer using residential histories. And so we really did a good job of looking at where people lived in the past in a relationship to their risk of developing breast cancer in the present. And then those clusters that we detected then generated hypotheses for us to pursue with relation to plumes and wastewater treatment facility located nearby. And so that's how we ended the talk. So most of the things that I do deals with GIS. And so for those of you not familiar with it, I just wanted to put one slide up that really summarizes what it does. And basically, it is a tool that allows you to visualize spatial data. And the map itself is usually made up of multiple layers. And that's really the the advantage to it is being able to take the spatial relationship of different features. And so here what we really look at and the work that I'll be showing is the relationship between where people live and the environment and the exposures that persist there. And so it's linking those environmental presence to people and to both the timing as well as the location to really determine exposure. So the story begins with a chemical known as tetrachlorothylene or PCE. It's also the chemical used in dry cleaning. Although this particular exposure that I'll be talking to you about was actually due to the fact that it was applied to pipes, the drinking water. So this chemical is highly volatile. So it's often used as a solvent when applying things like liners. And you usually believe that if you leave it there for long enough it will volatilize. There are no health effects. Most notably, it is a probable human carcinogen, both by IR and EPA. And it is commonly found in superfund sites. But the reason that we were interested was that it had been applied to a lot of the piping in Massachusetts drinking water pipes. Because of thisbestos cement that was used, people complained about this taste. And so as a solution they decided to put a vinyl liner on these pipes and they used PCE as the solvent. And so this is done for a lot of the pipes that were being installed and replaced from 1968 to 1980 throughout the state. What happened though is it was only done to those that people really complained about or that needed to be replaced. And so we had this really irregular pattern of exposure. So here the red pipes are those that have the vinyl lining. And the blue ones are other pipes. And so you can see that your neighbor across the street may not have actually been exposed even though you are. And so it's not just a proximity measure of exposure here. It's really knowing where people were living and when they were living in relationship to when these pipes were installed. So essentially what happened is they applied the liner and they didn't let it live long enough. And so after several decades, in fact, they started seeing this chemical show up in testing the drinking water. And so then immediately they began to purge all the pipes. But by then it had been several decades of people drinking this water. And this also temporarily coincided with a large increase in breast cancer risk. And a particular part of Massachusetts referred to as the apricape. Are people familiar with Massachusetts at all? Some of you. So Massachusetts has this tip called Cape Cod. And it's really basically it's a sandy beachy touristy area. And so the apricape is made up of these five towns. And most of the pipe with this PCE liner was actually installed in this area. So we've really focused our investigation of relationship between PCE and with breast cancer in this particular area. I lived in Palma for one year. Did you drink the water? I get the question all the time. Like, did you drink the water? Yes. And the pipes were in from 1968 to 1980. So what happened next was, well, they decided that they would do a couple of population based case control studies. The first one was 1983 to 1989 that they ascertained breast cancer cases. And then they followed up when there was more funding with another four years of data. So the true length diagnosis period for breast cancer cases was 1983 to 1993. And so one of the advantages is that with these types of study designs, you actually interview people. You collect information on risk factors that's really pertinent to determining if something like medication or hormone use or reproductive history, which are all really well known risk factors for breast cancer, is actually the cause of the disease and not drinking water. But in addition, they collect information on how much drinking water women consumed, how long they showered for. Because it's a highly volatile chemical, it's also important to know how hot your showers were. Because as you're showering, volatile chemicals will then enter the air. And so that's another exposure pathway for you. And so all these information will gather from questionnaire data. In addition, what we really had that a lot of registry based studies don't is an extensive residential history. So going 40 years back prior to diagnosis for cases or an index period for controls, all the addresses were collected and geo coded. And so people aren't familiar with that term. Essentially what that means is you take a street address. And in relationship to what's known about where coordinates are, you actually assign an X like coordinate or a longitude and latitude to people so that you can measure relationships spatially. So we have residency for all addresses for when they started, the year they started living there and the year they moved away. We also knew what their drinking water supplier was. So in this particular area, it's not uncommon to have private drinking water wells in your backyard. And so we know which women were having their own water supplied through private water or who were actually being supplied by public drinking water systems. And so that's important when it comes to knowing if you're actually being exposed to these pipes, because these were just public drinking water pipes. In the end we had 1480 breast cancer participants and that ended up giving us 2400 addresses. So really it's not just about the address of diagnosis again. We wanted to take a cumulative exposure going back to 40 years looking at all those addresses. So information was not just about the women but also about the pipes that we needed. We had the locations of where the pipes with the ACVL lining were as well as the years they were installed. And so then we could compare those years to the years women lived there, which is really important. We also had an idea of how much was initially applied then as far as the solid goes. As well as the rate at which this PCE was leaving the pipes and entering the drinking water. And so we were able to also use town land parcel maps so that we could geocode people to those parcels. And we knew approximately how many people were using this water. And so the trick really came in determining the direction of the drinking water flow. And that's a bit complicated. So essentially if you look at this house here in Yellow, if the water is coming from the north through the blue pipe, which doesn't have ACVL or the PCE, then they wouldn't be exposed. But if the water was coming from the opposite direction, where it passed by the red pipes, which is the PCE contaminated pipes, then they would be exposed. And so there was a lot of room for exposure misclassification. And this is a term you guys will probably hear throughout your study. But it's really when you think somebody's exposed, but they really weren't. And there could be various reasons for that. But this is one of the ones we were most concerned about. Is that if we didn't in fact take into consideration the appropriate flow of the water, well then we would be saying some people were exposed when they really weren't. And so in order to determine which direction the water was flowing, we used what was freely available, developed by EPA. It's referred to as an EPA net software package. And essentially, given the pumping rates and the amount of people on these different pipes, you were able to determine what direction the water was flowing in. And so although this is really used more for disinfection byproducts and other measures of drinking water quality, we were able to adapt it because of its open source to this particular exposure scenario, which was pretty innovative. And so in the end what we had was we knew based on the flow and the direction of the pipes what we were looking at as far as concentration of PCE. And so in the end, we calculated for each person and each address and then summed over time what their PCE exposures were. And then we used that in our regression model for all the breast cancer participants to see if there was a strong association between drinking PCE-contaminated drinking water and their breast cancer risk. And so in the end when we did this and as well as looking not just at their drinking water but also their shower and habits and things like that, it was actually one of my first dissertation papers was incorporating all their behavior patterns. And it turns out that, you know, not too many people take cold showers. So there wasn't enough variation in people's personal habits to actually make that addition to the model probably worth the time I spent doing it. But needless to say, until you do it, you don't know. But in the end we did see moderate associations. And the problem is that they weren't really strong enough to explain all their cancer risk. And there was still a lot of community pressure to really know and understand why this particular area was seeing such high risk. And so given the valuable data that we had in these case control studies, the fact that we had, oh yes. How high was the risk? It was about 1.4. Yeah, so it was odds ratio of 1.4, 40% increase in odds. People who were drinking this contaminated drinking water. And we looked at different latency periods and different lengths of exposure duration. And it did get stronger with longer duration as well as with longer latency. Breast cancer is one of those cancers which does have probably between 10, 20 year latency. And the word latency basically means the timing between when you're exposed and when the disease actually becomes apparent. So it's not that an exposure that occurred the day before you were diagnosed could actually probably be biologically relevant. It's less important than information collected 10 years prior. But again, we had all this valuable information on risk factors and where people lived, especially for me, and that's what I was most interested at the time was looking at how things spatially were relevant to etiology. And so what we did next is we took that information and conducted a space time analysis using the residential histories. So this is my favorite picture. This area of Cape Cod is really, it's unique in so many ways, not just because of this PCE exposure, which is pretty much unheard of in the rest of the country. But it's basically an environmental disaster area. There's a couple of different superfund sites. There's a large military reservation there. There was also a lot of pesticide use for cranberry bogs. And the problem is that the sandy soil, anything that sort of percolates in with the rainwater manages to get into the drinking water because there's one large aquifer for that area. And so then it all goes right back into your drinking water. And so this is a picture of a glass of water taken near the military reservation with enough jet fuel in it to light it on fire. And so for the longest time really this area, they were worried about the military reservation. But there was obviously other things going on with the drinking water and other sources of contamination. And so we wanted to also look fully at the entire area, not just at this one particular area. And disease mapping is a common way to do that, to take data that you have existing and do sort of exploratory analyses with that. And it's often done with cancer registry maps. As I mentioned before, there are some limitations. One is that it does ignore latency, meaning that it uses the address of diagnosis. But the other thing too is that they like to aggregate their data. And so they do that usually by just mapping to go rates in each town. But if you're thinking of an environmental exposure, it's not going to stop at the town line. And so really you're not getting a good sense of where the exposure is because you're having to aggregate. There's also what's known as spatial confounding. And that results from not having all of the appropriate risk factors collected. And a lot of times in cancer registries, you don't have that in some reproductive history. And so what we thought, well, we could take the data that we have and apply what's known as nonparametric methods to it. And so this is probably the more technical, mathy part of the talk. But essentially, I'll take you through it. But what it means is it's sort of a method that allows the data to give its own sort of its own shape to it. You're not fitting a straight line through anything, you're just actually letting the data smooth itself. And so for example, if you take the sine data here, you have on the left side what's known as an ordinary linear regression. So many of you have heard of y equals mx plus b. You're sort of fitting a straight line with the data. And that's not always appropriate. Your data is not always literally related to it. However, on the right side here, it's known as a smoothing regression here where you actually allow the data to take its own shape. And you do this by taking a window and almost fitting a straight line. But because the window is so small, those lines end up looking smooth or not straight. And so there are a couple of important things to note though about this smoothing is that if your window for smoothing is really too small, then you end up catching all the noise. And it's really a jagged smooth, which is kind of feeding the purpose of smoothing. But if you have a larger window that's just the right side, you can see that there's this trade-off between the bias and variance of your data. And so it's really the question when it comes to research methods is okay, what's the appropriate smoothing of that window? So the model itself looks like this. This is now your odds of having disease. It's known as the log odds. And you guys will learn that hopefully in your second year here. And the right-hand side looks like a typical regression where you would add a lot of your risk factors. But the one difference is that you're also adding a smooth term. And that smooth term is the exposure. It's your location. It's where you have your X and Y or your longitude latitude. And essentially what you're doing is you're fitting a plane or a surface to your data. So areas where there are higher risk is more of a mountain or a peak. And areas with lower risk turns out to be more like a valley or a low depression area in your surface. In addition to being able to account for location, you can also include all your known risk factors, which is really an important difference in a lot of other methods out there. So you're not only just looking at the location, you're also including the ones that you know are known risk factors like age, smoking, drinking, things like that. And then also important is that we can test for whether this effective location is statistically significant, which is important because you don't want to just say, oh, there's a cancer cluster here without being able to really support that with your hypothesis testing. Does anybody have any big picture questions on this moving forward? So what really happens is you take all of your data you've collected and you put in your model and here's where all that geocoding comes into play. And then basically you then take this model you've constructed and you predict it against all the possible coordinates or locations on your map in your study area. And it generates for you the smooth surface of risk where you see low values and high values. And then with GIS you can map that using colors so that the lower values are blue, the higher values are red. And so you end up seeing this map that looks kind of like a rainbow, but the areas with elevated risk are the warmer, hotter spots. Maybe this is a good time to ask this because I was thinking, for example, this chemical is binding California primarily because of dry cleaning. Right. So how do you draw a line between risk factors that may be relevant, may not be relevant inside the location? Right. So here because it wasn't banned we did include it as a known risk factor. We included it in our model and adjusted for it. And then we had the timing of it. And so it's really just looking at this particular case. If it were banned moving forward then there would be no exposure. And so that that timing would not be relevant anymore. But now we've sort of studied the PCE aspect of it and we knew that it wasn't the only explanation. And so controlling for that or including that in our model but then also including this term that really captures the residual risk or what's left over after you've accounted for what you know to be risk factors is really what we were most interested in. And so that's the exposure. It's just the relationship with space. And it was just going to show us if there was a pattern. And you know for some of the cancers that we looked at and some of the time periods we looked at, there wasn't any spatial variation. There was really nothing going on that differed from one part of the study area to another after you adjusted for things like age and smoking. But for breast cancer we did actually see a lot of spatial variation that even after considering things like reproductive history, PC exposure, other things that we knew or risk factors still persisted. And so we wanted to really study that further. So a couple more things. Again we are looking at odds ratios in this map. And so it's relative to the entire study area. We're using as our reference or our common group the entire average of the study area. And so just as a review an odds ratio of 1.5 means the risk of the breast cancer and that particular location or point is elevated 50 percent above the whole study area. And so the range of odds ratios was about 2.25 to 2.5 and the lower would be blue and the higher would be red. And so you'll see some areas that are white and that's areas in this study area that happen to be conservation land and there was nobody living there. So we excluded that. So these are a series of maps from the results of our space time analysis. And essentially what you see here is the scales are all the same. And so here we have this really elevated area of breast cancer risk. And there's also areas of decreased risk. And this is how it changes over time. And so the black bars here on this show beginning in 1947 and going to 1989 what the risk looks like over time. And so you can see the patterns do change. And this does support the fact that there is this latency period where the closer you sort of get to diagnosis the less is actually going on in relationship to the environment. But most importantly is that we had a couple of areas. One this is where the military reservation was and there were already ongoing analyses of that. And we also did see in this area some levels of increased risk. And so those are the ones that we were interested in exploring further because nobody had really looked into that. And so I just. I'm just curious which confounders are adjusted for. So this is. I mean you sort of hinted that there's a long list. There's like 10 of them. Okay. Age, all the reproductive history. Okay. Parity, history of breast cancer, smoking, drinking, PC exposure, hormone use, SES indicators. A lot of them weren't spatial confounders though. So that's the other thing because to be a confounder you need to be associated with the exposure as well as the outcome here, the exposure's location. And so really a spatial confounder needed to be spatially, you know, aggravated in certain areas. So for age you can consider that that would happen if you had a lot of elderly people moving to a particular town. But other things like the use of hormone replacement therapy was not spatially varied. So although it's a risk factor and other things it wasn't really a confounder here. But we included them and then, you know, saw that they didn't really change the pattern at all. Any other questions before I show you next slide? So I know obviously they read Zones of Concern. How come, why you compelled to try and explain why? Yes, that comes next. So first let me just, this, you know, one of the highlights of my career here. So sad. So what happens is for each year you take this time period and you smooth the data and then you can generate a map. Well you take the maps and you put them next to each other and this movie maker software and really it just shows the movie over time of how the risk does change. And so it's just visually really almost startling to see how risk was so much higher, you know, 20, 15 years back and how it changes and gradually, you know, you don't see it as much anymore. And so yeah, we wanted to know what's going on in some of these areas. This sum really does sum it up fast here, where we have for our no latency, which is all of our addresses up until diagnosis really nothing going on. It's a flat surface. And then here this is with our 20 year latency, we see a very significant p value, which means location certainly doesn't matter here. And then to be specific, we're looking at these two areas. And again, I've explained this one being the military reservation, but this one here is the one we wanted to follow up more. And what we did is we overlaid that with existing maps that were generated through the conservation area in Cape Cod of known plumes in the area. And so the black contour lines now represent areas of statistically significant increased or decreased risk. And you can see here that these two areas in the far right here are quite similar in both in shape and size. And so this was really our jumping off point. We saw that there was this relationship. And of course, this doesn't mean that it's in fact caused by one thing or the other. But it did sort of, in our minds, tell us, well, this is something that needs to be explored further. And so these are plumes. The ones here in dark purple are the plumes from the military reservation. But the ones in more of a pink are ones from other sources. And those sources include landfills as well as a wastewater treatment facility. And so that's where the story goes. Next, we have the spatial co-location. But that's obviously not enough to say that you know something was going on. What we needed to do is go back to the questionnaire data, find out more about their drinking water and what their source was and what companies were supplying them as well as where the drinking water wells were located in relationship to these plumes. And so the first problem though was that we didn't really know what was going to be in the plume. Unfortunately, because it's a retrospective study and where the exposures really were occurring back in the 60s and 70s, we didn't have any measurements anymore. They didn't take any at the time really. They just didn't know to you. And so we couldn't say, even if we had an idea, what would be in the plume or what the actual carcinogen would be. So right from the start, we were limited in the fact that all we could say was that effwint coming from a wastewater facility or leaching coming from a landfill or whatever it turned out to be was related to breast cancer risk. But unfortunately, we could never say for certain, Jesus, that it was a specific carcinogen. And that really is, you know, it's almost bittersweet for people who are more confident now that was drinking water related to breast cancer, but still don't have any idea what the culprit really was. But wastewater does have a lot of mammary carcinogens, including PAHs, pesticides, pharmaceuticals, and the wastewater treatment facility that we focused on collected commercial waste as well. And so then you don't even know if you're dealing with dyes or other really harsher chemicals in addition to the normal pharmaceuticals that you would expect to be in wastewater. So there were several potential sources of contamination to the drinking water. And these included five landfills, one for each town, as well as two municipal wastewater treatment facilities. And so the wastewater treatment facilities up until not too long ago did what's known as primary treatment. And so they filtered out the large parts of effluent, but all the liquid went back into the groundwater as part of the recharge for drinking water. And this was common, this was appropriate, this was legal. But when you think about it, it's just like all your legal, your liquid waste is going right back into your drinking water. So now there's much stricter standards in secondary treatments. But at the time, this is all that was done. And so you can see how it would be plausible to have something that was existing in the wastewater treatment effluent mixing with your drinking water and leading to health effects. So again, this is a map of the the contamination sites we're talking about. And so we have landfills here as well as this wastewater treatment facility in this one here. The Peacock Commission, which is that resource that we alluded to earlier, they had already had concerns of the drinking water. And so the US Geological Survey had done some impact studies. And this was done in 1993. So these red lines delineate plumes in 1993 that they believe may be the direction and the shape from these potential sources. And so again, this one here from the wastewater treatment facility was really the one that we were focused on. Oh, that's weird looking. Okay, you don't know why that happened. But so here you can see in yellow the public well locations. And really we were interested in, okay, well, it's not enough to just have the plumes there, but people needed to be getting contaminated somehow or exposed somehow. And so really, we needed to see a location of a well that actually was within one of these plumes. And so there were several different contamination sites. The only ones that were actually impacting public drinking water wells was this wastewater treatment facility here. And so this is just zoomed in. You can see that there are two wells here that were within the plume. And so knowing what their drinking water was, we could then link those particular wells to a particular drinking water company and know which women in our study were actually being supplied drinking water from that company. And so they're shown here in orange. And so you can see that these are women that are actually living outside of the plume, right? So it's not just about being within that plume, because the pipes that supply this drinking water, where the wells are located in the plume, are actually outside of the plume. And so just doing a simple study of whether you're in the plume or out of the plume is your exposure measure, but again, lead to what's known as exposure classification, because you're missing the people who are being supplied this drinking water through drinking water pipes. And so here we have, in addition to those in the public, the women that are reported being on private drinking water. So they had in their backyards their own private drinking water wells. And these were women that we knew would be getting water from living within that plume. They were the ones that were living in the plume, that's where it mattered. The other important part of this though wasn't just the spatial element, it was also the temporal element. We needed to really sort of cross-classify where the plume was over time with where women lived over time. And so the plume began in 1937 with the operation, the facility operating. And as I mentioned, it was primary treatment. And so we have records from the town reports in this area, the amount of sewage that was actually released and processed each year. And so you can see over time that that does increase substantially. And that's largely due to the fact that the population in this area is growing rapidly. And so it's not just even that that the plume is moving, it's also that the amount of effluent within that plume is also growing. What was the gap in that? Yeah, fire, I don't know. There is this gap in data here. And really it's just the fact that the town books didn't have anything recorded for it. And the end we didn't, because we didn't know what was actually in the effluent or what chemicals were there, we couldn't really use much of this information besides really doing more of a proxy of the volume. So the technical modeling aspects are summarized here. We use what's known as a three-dimensional mod flow groundwater model, and this is developed again by the US Geological Survey. And the importance of this is that it takes into account not just where the plume is moving in the x direction and the y direction, so basically across from you, but also in depth. Because it's really important to know that the plume or the contamination reached where the pumps were actually dug down to you. And so that's why we went with the three-dimensional model. We had information, a lot of it had already been collected by the USGS on the geological properties, but we also had to go back and incorporate historical information. Because the USGS and the plumes that you saw in red back then were all from 1993, we really need to go back in time and say, well what did the plume look like, say 1967 or 1977, when these women were living there at different time points? Because it could be that the plume hadn't reached the public water wells yet, or the private residences with the private wells yet. And so this is all really this matching of time that we needed to do. And so the model was run for each year beginning in 1937. And given all the pumping rates as well as the location of ponds and things like that, we determined that the public wells actually only first were contaminated in 1966. So this is about 30 years after the first drop of effluent was released into the environment to actually reach the drinking water wells. And so that's a substantial amount of time. But also important to know is that that's the first of the effluent. So that's where all the commercial waste was still in there. That's when really it's the older chemicals. But unfortunately we don't know what those are. We don't know anything about retardation. So we really just had to assume that the effluent was moving as quickly as the brown water was moving. We also assumed that the participants that were served by the drinking water were all equally exposed. It's not a bad assumption to make just because of the way that the water is stored. They're all stored in what's known as common sandpipes and so it's mixed together. And so we're just assuming that all the women scattered throughout this pipe network were receiving the same concentrations. But most importantly we had their residential histories. So we knew when they moved in, when they moved out. It's possible that a woman could have moved in and moved out well before the water was contaminated. And so this was really the best way we had in order to make an accurate assessment of exposure. And so here's another little wee clip of the progression of this plume over time. And so this is the wastewater facility. And then the darker the purple gets, the more effluent in that particular cell of the model. And you can see that eventually it does sort of spread out and reach these private wells. But it does take a while to actually get there. And so when it could have you know been misclassified as women immediately in 1940 being exposed, we determined that it was really much later. Not until the 70s that it really became an issue. What we also had to do or to the best of our efforts was to validate the model. This is really important. You guys will hear this over and over again, I'm sure. But it was hard to do that because again we didn't know it was in the water. So there was no samples available or collected to really be able to use as a validation method. What we did know is that our model was based on the USGS model and did in fact give us similar results to theirs. We also had what beginning in 1972 the safe water, the Safe Drinking Water Act was passed, which mandated collection with at least nitrate samples. And nitrate is really a good indicator of wastewater in the drinking water. And so in starting in 1972 we did have nitrate samples. So we were able to validate the model after that and we did see a relationship between the levels of nitrate and samples collected at these wells as well as the model where we predicted that nitrate or effluent was reaching it. So this did support our model results. So for statistical analyses we did a couple of different exposure classifications. We first looked at people divided by the duration. So how long did they live there? People obviously that were living there a lot longer were consuming a lot more of this drinking water. We also had over the entire residential history a cumulative measure, but then we stopped at certain time points to allow for latency so that we could also consider the fact that drinking a glass of water the day before you were diagnosed isn't going to be that relevant. We had a common unexposed reference group of 700 controls. So these are women that did not have any drinking water exposure whatsoever throughout any of the residential histories. And then we had 533 cases. So we controlled for age, vital status, family history, personal history, breast cancer, age at their first birth, education, race and study at war age. So for the results you can see on the red is the latency periods and I'm just going to show you a couple of latencies, 0 to 15 years. And then here is duration. So this is living on the cape for less than five years or equal to and then for greater than five years. And so we have the number of cases and controls as well as our adjusted odds ratio and 95% confidence intervals. So you can see with no latency that there is a now relationship here especially with less than five years and non-significant although slightly elevated risk for greater than five years. But then when you look at with a latency period of 15 years as well as with a five year or more duration you have statistically significant risk of 1.8. And this actually got stronger with 20 years but because of date limitations it was really unstable. So what I mean by limitations is that not only do you have to have lived there 20 years prior because of latency you also had to live there for five years prior to that in order to be part of this longer duration category. And so we just didn't have a large number of women who were actually fitting that residential identity. They weren't there for 40 years or very few of them were. So it'd be interesting to look at this today with new batch of women to see if this still holds but that actually involves more money. So in conclusion we actually did see one of the strongest relationships between environmental exposures and breast cancer in this particular cohort through all the studies that we had done and so that was almost an odds ratio too. And it only got stronger with longer latency, longer duration. And again there wasn't historical water samples available besides that of the night traits but we're fairly confident that the model is predicting pretty accurately as far as timing of exposure went. There isn't this possibility that there are other things in the water and that's you know something that we do lead to in our paper too. There was an airport nearby that could also have been spewing jet fuel into the drinking water. So there could have been other special confounders that were also playing a role in this. But the bottom line is that it was looking like something in the drinking water at that particular time didn't impact the risk of developing breast cancer between 1983 and 1993. And so as a follow-up to this a little bit is that there's been quite a bit of public response and it's all just a time issue but the paper came out in February and then it was quiet, there was nothing. And then the elections happened and one of the items in this particular time was the expansion of the wastewater facility. And so I get all these emails and phone calls and bloggers about oh you have to you know take our side and say that the wastewater treatment facility is bad and you know expanding it would be horrible for you know the health of the community and the drinking water. And so it became really an issue and you guys will learn this in public health about translating your research and how that's best done and really trying to express a complicated issue of timing of exposure. So to say well you know I can't speak to what's going on now is basically my bottom line to them because this was historical right. So the cancer risk here that we're talking about was from 1983 to 1993. And the exposures again were really the ones that occurred with the beginning of the wastewater treatment facility and who knows what was in that effluent. And so I felt really bad just trying to explain to people that well it was bad then but I can't say that it's bad now because I don't know and that's not the answer they wanted to hear unfortunately. So it was both sides being mad at me and like I couldn't say one way or the other because really it's not a good idea having drinking water wells a mile and a half south of a wastewater treatment facility. Like it's never a good idea but to say that this study that we did suggests that there's health effects today was also misleading. So it was definitely my first exposure into the world of politics as far as a community level goes and it was interesting I have to say but I prefer doing the research part of it and letting somebody else talk about why or why not you know it can be used to their political advantage. So it was an eye-opening experience in our respect. So just to wrap up here I wanted to acknowledge that this work was funded by the Superfund Research Program BU as well as a grant that I had from the National Cancer Institute to do some of the space-time work and also like to acknowledge co-authors. Lisa Gallagher was my graduate student she's recently graduated and is now at University of Washington doing her postdoc and then colleagues Tom Webster and Ashton Ground Janice Weinberg who've all been working together on this sort of spatial epidemiology group that we have at Boston University. And lastly I'd like to thank you for being great on it and if you have any questions please don't hesitate to ask. Last thing too is that we have this website here with more information I don't know if you guys can see it from there but feel free to take a note of that and what I'm sort of moving towards now is incorporating Google Earth into our space-time mapping because it just looks cool yeah. I know you mentioned towards the end that you had people calling you and wanting to take their side but in general what is like the people living in this area what is their reaction I mean are they scared to drink the water? You know I got that question to you and I honestly answered well yes I would drink it down because there has been so much more done in order to protect the drinking water. I also had a town counselor ask me well why are you doing this if it's not relevant to today and I was taken aback by that because there are women that have suffered with this disease and you know at the time they you know really want to know and for those that are still living and hopefully there are many you know it's still important to them to understand why breast cancer risk was so high then and the risk isn't as high anymore it's still slightly elevated compared to the state but not as high as I've had by that time period so there's really this question for a lot of public health officials why breast cancer risk was you know elevated and I think it was important for those women for the participants I mean first and foremost we do this for our participants you know they're the ones that are looking for that explanation and so to be asked you know well what good is it now it was slightly I don't know offensive I guess but there are lessons to be learned now to you you obviously do you want to continue to protect your drinking water and this it's still true that they don't really have a good sense of what's in the drinking water there's a lot of emerging contaminants out there so this whole other research area that I work on deals with um chlorinated compounds and you know scott artel and they're working with that as well but there are just lots of things that are released into the drinking water through pharmaceuticals through you know flame retardants things like that that are getting into the drinking water and there's very little screening for them in general you know it's not something that needs to be looked for yet and so little known about the health effects too you know on the other hand but it's sort of a precautionary principle tale you want to try to limit what you're drinking even if you don't know if it's going to make you sick you'd rather just get as clean as possible drinking water as you can and so that was really my take home message to them but you know they've shut down the pumps you know the the contaminated ones in that plume kind of indicated they are they doing special treatments those are they shut down this they are continually monitoring them you know and they're meeting the standards okay so the nitrates they've sort of dropped but again they're only a proxy right so but that probably indicates the waste water plumes that are sort of passed by the wells now they have and so they've also changed pumping rates which affect the direction you know and so I think that they're drawing more water from other wells and so that's really moving the course of the plume the water flows to where the pumping rate is higher so it's actually changing the direction of it a bit too they didn't shut down one of the wells not because of the effluent but because it was being contaminated by some other VOC and so again this is a soup of stuff going on and it's hard to tease things apart and one of the other things you know I thought about is this is a really great chemical mixtures problem because I'm sure there's multiple exposures going on here the problem is is we can't quantify any of them and being retrospective it's almost impossible to do now and moving forward you know there's probably still plenty there to measure but it would be you know I think difficult to acquire funding based solely on that yes yeah you know for private drinking water what they say is you know you can get these companies to come in and test most of the time what they're looking for when they do that is bacteria or metals and things like that and a lot of them just a lot of labs special locally labs don't have the capabilities to to test for the fluorinated or the polybrominated compounds and you know it's just something that's not done regularly now but you get your water report you do but again those aren't there you know these emerging contaminants are rarely measured in that unless they're specifically being looked for and so in this particular study area we have some collaborators who are a local nonprofit company dealing with breast cancer issues and so they've done they have collected water and sent them to the few labs out there that do look for these emerging contaminants and have detected them and I'm sure they're detectable throughout the country EPA's also done a similar analysis of the chlorinated compounds looking at how those are emerging in drinking water but you really do have to find a lab that's qualified to measure them because it's not the easiest thing to do in drinking water so the things you get in your drinking water report is mostly just the metals and the bacterial loads and and the usual sort of concerns there's a lot of VOCs there's a lot of you know known carcinogens that are detected but the emerging ones haven't quite made the list yet they're over 80 000 industrial chemical license for use in the U.S. and we don't really understand the toxicity of a couple hundred of those it's probably about a hundred are mandated for testing in drinking water so is bottled water a better option? I mean right now this plastic is just leaching in drinking water right so it's like you're not safe for anything and the waste that used to be too that they've mandated less testing of bottled water than public drinking water you've got to do that but you know they're probably only half as long the list of chemicals that have been mandated for testing of bottled water and you don't really know where they're getting the water either so it's I think the best bet is right now granulated active carbon filtration so your rid of filters or your fridge filters really do remove a lot of the metals and even these perflornated compounds that remove them so it's really one of those things if you can maintain changing your filters it's probably the best safest bet for drinking water at this time point it's different it's a different chemical structure so fluoride's good for you whereas teflon is not something you should be eating so it's it's a different chemical structure it's more about it's got a long organic carbon chain with the fluorinated compound at the end so well just recently do you know that they're going to stop adding fluoride? Right I've seen that yeah that's not really my my field of expertise although fluoride's been a contentious issue for a long time so I actually have a colleague who's done a lot more work with fluoride but I haven't been quite following it the chemicals that I'm referring to it's one of the chemicals you use in the manufacturing is teflon or for scotch guard and so it's something that's in a lot of consumer products but because those products are now getting back into the waste stream ending up in landfills and then leaching into drinking water that's how they're getting into our drinking water so it's very circular any other questions? I'm not sure what the population distribution is but these blues are reduced risk what would be the and they're just reduced in this particular area they could still be elevated compared to the state right and so you know but they did have some lower areas less than one odds ratios and you know a lot of them there's several different public drinking water areas most of them are from this aquifer but some of them are do you have a question? I was just wondering in the wastewater plume area were there pcd treated pipes they were they were controlled for so it wasn't the risk that we saw after it wasn't due to that but yeah there are a few of them yep it seems to me just from what you're saying because you have this sort of historical well you have this historical presentation but we're missing sort of some of the crucial additional data that um I mean as both ideally same examples would be great but I don't know where it's based but but actually measuring doing more elaborate measuring that would be what you recommend from this so moving forward exactly even if we don't know what we're going to get 20 years from now you have it on hand yeah right I mean that's what we're trying to do is archive as much as we can now and and then in the future you'll have it for looking at you know relationships with health outcomes that you now know you know our result of that and I thought you know maybe we could do sediment cores there are lots of ponds maybe there's a way of actually going back and measuring you know these chemicals and in other media besides drinking water and then correlating it back but it would you know it's still just an idea right now and it could take a little bit of work to correlate that back but who knows I mean it might be archived there in some mucky pond somewhere what was happening 30 years ago so I haven't completely dismissed the idea but it takes a lot of money to do sediment cores so we'll see any other questions so I it's going back a little bit more technical quick but could you comment on the the issue like when you're developing your spatial models with your planes that you are looking at every point which is you know geography in these days but your population densities are not right so the type of smoothing we use is sensitive to density because it's not given a uniform area it's given a number of people so the smooth will go out as far as it needs to collect data until you've reached that minimum number of data points not until you've reached that geographic area delineation so in areas that are really highly populated you have really a smaller window of data used for your smoothing but areas that are less densely populated then your smoothing window or the area that you look at to collect your data really goes out much further and so it takes into account this changing of population density so that you don't have fluctuations in your surfaces due to the fact that there's only five people living within a square mile so does that express itself in a change in the confidence interval it does yeah it's in the confidence intervals so we see it in our permutation tests as we recreate our distributions using Monte Carlo simulations so it's we're also exploring now to kind of evolve our methods looking at splines because that uses a locally slated smoother but you know we think that splines might be a better way to incorporate both the space and time elements into one model but right now we're taking time slices and then applying a spatial model to it so it'd be nice to incorporate both terms into the same smooth so it's exciting where we're going with it but again we need to have the right data sets to apply it to yeah right so I mean for exposure being the term of in terms of space yeah of the risk in that area you are going to you know different sizes of areas as far as that goes but the relation to the plume it's a little different but yeah for exposure misclassification I mean it's hard to really think of exposure misclassification in terms of this spatial smoothing but it really does look at a wider area and so that is one of the limitations we've done this analysis with the nurses health study which is looking at a national cohort of women and there are several states in the US that just don't have very many women so what we do is we actually exclude those from our predictions because of the fact that there is so much explosion misclassification you're taking women from like both ends of the country to predict from the middle and then that's not really a very representative measure what's going on in that location so yeah in terms of that because this area of the density wasn't as varied we did have fairly uniform we weren't worried as much about that except in the areas of conservation land things like that where there were only a couple of people you know scattered nearby so those we did exclude as well any other questions okay thank you thank you