NCI NewsCenter
NCI Budget Data
Visuals Online
NCI Fact Sheets
Understanding Cancer Series
Chart of cancer statistics
Each year, the NCI, along with various other health organizations, compile a report that statistically assess the nation’s progress in the war against cancer. This year, the report focuses on how an aging population could lead to a doubling of the burden of cancer in the United States by 2050. The report also examines current incidence, mortality and survival trends. Because a number of changes occurred this year in the way that statistics were assembled and reported, BenchMarks talked with two of the NCI authors to get a better handle on what these changes mean for cancer in this country.
Brenda K. Edwards, Ph.D., and Lynn A.G. Ries, M.S., of NCI’s Surveillance Research Program and home of the Surveillance, Epidemiology and End Results (SEER) Program, discuss the 2002 Report to the Nation, statistical methods, and the various statistical changes that occurred during this reporting period:
Why is it important to age-adjust cancer statistics and why can’t we just look at overall rates?
Ms. Ries: Age adjustment is a method used when comparing rates among two different groups in case they are different in their age distribution. For example, if we were to compare two geographic areas like Florida and Alaska, where Alaska has a much younger age distribution, it may look as though Alaska does not have as much cancer mortality as Florida, but if you age adjust, they have almost identical cancer mortality rates.
Dr. Edwards: Because the rate of cancer is very much influenced by the age of an individual, when we’re comparing cancer rates in different groups of people, or as Lynn said, groups of people in different places over time, it’s important that we look at what’s happening with the cancer rates of different age groups and find a way to summarize what’s happening in each of the age groups so that we’re looking at those patterns, not the fact that the two groups may have different proportions of individuals at different ages. So we’re trying to eliminate the age composition difference in the groups and instead focus on a way to summarize and compare what’s happening in each of the age groups.
Ms. Ries: If cancer rates were the same for each age group, then probably age adjustment wouldn’t matter, but the problem for many cancers is that cancer is rare at younger ages and is more common in older age groups. If you have two different groups that vary by age, a younger group will look like it has a lower rate and an older group will look like it has a higher rate, when actually it has nothing to do with the actual risk of cancer.
Why do we look at rates and not whole numbers?
Dr. Edwards: Because it gives us something that’s comparable in terms of the risk of cancer. It’s a measure of how many people either are diagnosed with cancer or die of cancer in any one age group over the number of people in that age group. So it’s important for us to look at those rates, those percentages, when we’re comparing groups of people. The count is important too, because that reflects the burden on the population – the number of individuals who need to be treated and cared for. But if you’re looking at the risk of cancer, and if we’re seeing any change in progress, we’ll see it in the rate.
Ms. Ries: You’ve got to keep in mind, if we go back to a geographic example, the state of Alaska has many fewer people than Florida and many fewer cancer deaths, so it would be unfair to say you don’t want to live in Florida. It’s because there are more people that you have more cancer deaths; the risk of dying of cancer is not necessarily higher. When looking at a racial group or a cancer site, it’s really dependent on how many people are at risk. The number itself doesn’t tell you enough, but the rate gives you more information on what the risk is in that particular group.
Why do we need to look at various age groups, such as young, old, the oldest old?
Dr. Edwards: We’ve already said that the cancer rate does increase with age and it’s very important to see what kind of patterns or trends are occurring in each of these age groups. Often what we see is a reduction in the risk of being diagnosed, and certainly of dying of cancer, and we see reductions in those at younger ages. It takes longer to see the reduction or improvement in older populations that have the highest rates. For example, in something like lung cancer, we started seeing a reduction in lung cancer mortality in men beginning in the younger ages and now we’re seeing it in the older men.
So you can’t just look at the overall rates?
Dr. Edwards: That’s correct. The overall summary we give you in age-adjustment is the one simple measure that talks about what’s happening across all the age groups. It is really a summation of the patterns in the various age groups in the various cancer sites. By looking at the level of detail presented in this report, it helps inform us and interprets the overall summary or the age-adjustment patterns.
Ms. Ries: Just because a rate goes up or down, doesn’t mean that it goes up or down for each age group. There are some mortality rates that are going down for the younger age groups but not for the older ones. In people over age 85 it may be going up slightly, so it (a detailed age group report) really does give the idea of exactly where the increases and decreases are occurring.
This year, for the first time, you’re adjusting to a year 2000 standard population. Why is that and why are you doing it now?
Ms. Ries: Our program began in 1973, and the 1970 population was the closest we could come for age adjustment, but now that we have data for 1999, it seems a little out of date to see what rates in 1999 looked like compared to a population in 1970. So starting with 1999, the federal government decided that it would be good for all the federal agencies to use a more contemporaneous standard, which is the 2000 age distribution. This allows us to look at the 1999 rates as though they occurred in a population of today.
Dr. Edwards: There are two main reasons. The first one is that many of the health agencies were using different standards for adjusting populations to standardize their reporting of various health measures. We needed to bring our reporting to where we’re all using the same standard because it was making it very difficult, if not impossible, to compare reports from one agency to another. So the first reason is that we would all be using the same standard and the reporting of health data would all be comparable. The second reason, as Lynn said, if you look across the various groups reporting health data, it was time to bring those age adjustment rates to one that reflected our contemporary health picture today rather than 30 to 60 years ago, which is what the other standards had been doing.
Previous reports have noted a decline in both overall combined mortality and incidence, yet this year incidence is classified as ‘stable’. Why is that and does it have anything to do with standardizing to a year 2000 population?
Dr. Edwards: Actually last year we began to notice that some of the rates were going up. We brought attention to the breast cancer incidence rates last year. We had also noted that other sites, particularly prostate cancer, had begun to turn up, but we didn’t see a significant increase at that point. So this year, the stabilization of incidence rates for a more recent time period is due to the fact that breast cancer has a statistically significant increase in the recent time period and we see an increase in the diagnosis of prostate cancer in men less than 65 years. Increases in these two sites, even with declines for other sites, leads to a stabilization of the reported incidence rates.
Ms. Ries: That stabilization is going to occur whether we use the 1970 standard or use the 2000 standard — it’s more that we’ve added data from 1999 that causes us to see this stabilization.
To what extent have rates prior to 1999 been standardized to the year 2000? Is all of the Cancer Statistics Review (CSR) data that went up on the SEER Web site on April 15 standardized to year 2000, but archives from previous years only adjusted to 1970? In other words, what is the extent of the data that has been retrospectively adjusted to year 2000?
Ms. Ries: All of our data are self-contained within the Report to the Nation, so that if you go back to 1973, it has been adjusted to the year 2000 standard and all of our data has been updated–it is self-contained and it is the one report to use. You would not want to take a figure, a particular number, from a previous report and compare it. However, in each report, all of the data are consistent and age-adjusted to the same standard. In the Cancer Statistics Review, we actually showed data from 1973 through 1999, and all of the data are age adjusted to the year 2000 so that you could get data points back to 1973 and you don’t have to go to previous volumes to find the data points.
And last year, you did this same adjustment to the year 1970 standard population?
Dr. Edwards: Right, because that was our standard in effect at that time. For most people, the two important messages are that this report is complete, comprehensive, and should meet most everyone’s need for the latest data and includes data for 1999 and updates for any earlier reports. The second point is that it will complicate people’s lives to try and compare any estimate with previous reports, but this should not be necessary because all of the data are in this report and the latest CSR.
Ms. Ries: Again, because cancer rates tend to go up with age, when you age adjust to the year 2000, many of the actual rates will look like they go up if you compare rates age-adjusted to the year 1970 with those age-adjusted to the year 2000. The main point is that if you age adjust everything you’re looking at to the same standard, then you can actually compare those rates. You don’t want to compare rates age adjusted to one standard with those age adjusted to another standard — it will look like an artificial increase in the rates, so what you want to do is compare apples to apples rather than apples to oranges.
Dr. Edwards: I think it’s important to note that age-adjustment is a method to summarize data that is occurring at many ages and at many ages across time. It reflects a method, a technique. It doesn’t alter underlying patterns at each of the age groups. That’s where the important information is, not the method we’re using for age adjustment. We happened to have changed that method; that’s why some of the numbers appear to be different, but it’s a function of our method for summarizing cancer rates across many different age groups.
When graphing trends, particularly for joinpoints (which indicate where cancer trends are shifting in time), are all of the data points adjusted to the year 2000, and if not, why? If so, does this mean all SEER data going back to 1973 has been adjusted?
Dr. Edwards: Age-adjustment to the year 2000 can have some impact on what the overall trends look like as measured by joinpoint, but the important piece is that we’ve added another data year and that too has an influence over what these trends, or joinpoints, are going to look like. Overall, use of the (year 2000) standard has had limited, or modest impact on the overall trends. In fact, we have looked at adjusting to the year 1970 standard and compared it to trends adjusted to the year 2000 standard, and there’s almost no difference.
We say that a year 2000 adjustment may make some rates appear to be 20 to 50 percent higher. What are the most glaring examples of this difference and why do the data for these particular diseases show such large differences? What cancers, if any, are barely affected by the adjustment?
Dr. Edwards: Adjusting cancer rates to the year 2000 standard gives more emphasis to cancer occurring in older people. Most cancers, certainly the top four (cancers), occur more in individuals over the age of 60 or 65, so those are the cancers where we’ll see the biggest difference.
And those are?
Dr. Edwards: Lung, breast, prostate and colorectal, as well as others we’ve noted in our report. Where the average age of cancer is occurring in someone over 60, 65, 70, those are the cancers we’ll see most being affected by the age-adjustment to 2000. It will have limited or negligible influence over those cancers that occur among younger individuals– for example, testicular cancer.
Ms. Ries: Though it may look as if the rates are going up by 20 to 50 percent, we’re not saying that an individual’s risk of getting cancer increases by 20 to 50 percent. What we’re saying is that if the population, as it does today, looks like it did in 2000, then this is the risk of getting cancer. Before, what we were saying was that if it looked like the 1970 population, which it doesn’t, then the risk was lower. It’s really looking across all age groups. If you look at age-specific rates, they are still the same, no matter which standard you use.
Dr. Edwards: And in fact, age-adjustment is a method by which we average or summarize the risk for being diagnosed with, or dying of, cancer, and those risks, by and large, in every age group, are going down. Because we’re using a method for summarizing across age that gives more emphasis to older populations, we have the appearance that this summary measure, the overall age-adjusted rate, is going up. The good news is that if you look at individual ages, my risk of dying of cancer compared to someone my age 10 years ago, is lower.
New prevalence data shows almost no change in the 8.9 million cancer survivors in 1999 vs. 1997. Was this number calculated using year 2000 standards, and if so, why isn’t there a change in prevalence estimates?
Dr. Edwards: Our prevalence of 8.9 million cancer survivors reflects an estimate based on some new methods that have been developed for obtaining estimates on how many people are alive today who have been diagnosed with cancer in the past. It actually differs from published methods used in the past. The previous estimate was based on data from just one state with a registry that went back 50 or 60 years. This new method is actually somewhat conservative because of the way in which we’re counting cancers, particularly cancer for people who may have had more than one cancer, and how we take that information into account. It’s actually a much better method because it allows us to estimate prevalence by many different characteristics of the population — not only cancer type, but age of the individual and how long they’ve been alive with the cancer. It will give us a lot more information for characterizing our cancer survivors and in planning for health services for cancer survivors.
Ms. Ries: One of the things we did in changing the statistical methodology is to only count one person one time. So a person who had both breast and colon cancer are both a breast and colon cancer survivor, but when we add up the numbers they would be counted for both cancers, which, while it sounds okay for each individual cancer, the overall number would reflect more people than are actually affected. Therefore, looking at our previous numbers, it seems we’re staying at 8.9 million survivors despite an increase in the overall population for another year. It would be a larger number if we allowed each of the cancer sites to include all of the people that had that particular cancer, but this time we decided to limit it so that the total number would reflect the total number of people that had cancer.
Where will the change from ICD 9 (International Classification of Disease 9) to ICD 10 be seen most prominently? In which cancers?
Dr. Edwards: Our report this year is having yet another change. The U.S. adopted a new standard for classifying death, the ICD 10 version. The consequence is that it actually affects all causes of death, and for cancer, as Lynn indicated, it has the impact of increasing the way in which we identify cancer as a cause of death. About seven-tenths of one percent more of the deaths will be identified as cancer with this new procedure. In addition, as she also indicated, we will see some changes in the specific types of cancer that will be identified as causing cancer. That’s due in large part to situations where we have a different way to count people who die with more than one cancer that causes their death; secondly, the way in which many people are reporting death where there’s metastatic disease.
Animation/Video
| This animation requires the Flash plug-in. If you do not have the plug-in, please click here to install. |  |
Text Transcript
The first segment of the animation shows the chart, Average Annual Cancer Incidence and Death Rates, 1995 to 1999, which depicts how the rate of deaths and rate of new cases of cancer have changed in the U.S. population for the period 1995-1999 for four different age groups.
The chart consists of two graphs. The first graph depicts the average annual cancer incidence and death rate for males in the years 1995 to 1999. The Y axis, the vertical axis, is labeled “Rate per 100,000″ and lists the values 0, 500, 1500, 2000, 2500, 3000, and 3500. The X axis, the horizontal axis, is labeled “Incidence” and “Mortality.” Below the X axis is a key explaining the colored bars used in the graph to represent patient age groups: a yellow bar represents ages 50 and below; a green bar represents ages 50 – 64; a purple bar represents ages 65 – 74; and a light blue bar represents ages 74 and above.
As the animation progresses, the yellow bar rises to 74.1 incidences per 100,000 and 21.6 deaths per 100,000; the green bar rises to 934.2 incidences per 100,000 and 343.1 deaths per 100,000; the purple bar rises to 2604.9 incidences per 100,000 and 1067.1 deaths per 100,000; and the light blue bar rises to 3210.2 incidences per 100,000 and 2013.3 deaths per 100,000.
The second graph depicts the average annual cancer incidence and death rate for females in the years 1995 to 1999. The Y axis, the vertical axis, is labeled “Rate per 100,000″ and lists the values 0, 500, 1500, 2000, 2500, 3000, and 3500. The X axis, the horizontal axis, is labeled “Incidence” and “Mortality.” Below the X axis is a key explaining the colored bars used in the graph to represent patient age groups: a yellow bar represents ages 50 and below; a green bar represents ages 50 – 64; a purple bar represents ages 65 – 74; and a light blue bar represents ages 74 and above.
As the animation progresses, the yellow bar rises to 113.7 incidences per 100,000 and 23.2 deaths per 100,000; the green bar rises to 802.8 incidences per 100,000 and 269.1 deaths per 100,000; the purple bar rises to 1527.4 incidences per 100,000 and 680.3 deaths per 100,000; and the light blue bar rises to 1931.4 incidences per 100,000 and 1148.3 deaths per 100,000.
The second segment of the animation shows the chart, Projected Number of Cancer Cases for 2000 through 2050, which depicts the projected rise in the number of cancer cases in the U.S. population between the year 2000 and 2050 for 5 different age groups.
The table shows that for the age group of 50 and below, shown as a yellow bar, cancer cases remain steady at approximately 200,000 a year for the period 2000 through 2050; For the age group of 50 to 64, shown as a green bar, cancer cases rise from approximately 550,000 a year in 2000 to 700,000 a year in 2050; For the age group of 65 to 74, shown as a purple bar, cancer cases rise from approximately 900,000 a year in 2000 to 1.5 million a year in 2050; For the age group of 75 to 84, shown as a light blue bar, cancer cases rise from approximately 1.25 million a year in 2000 to 2.25 million a year in 2050; For the age group of 85 and above, shown as a red bar, cancer cases rise from approximately 1.4 million a year in 2000 to 2.65 million a year in 2050.
Audio Clips
- First Lynn A.G. Ries, M.S. and then Brenda K. Edwards, Ph.D. (both from NCI’s Surveillance Research Program) discuss the age adjustment to a year 2000 standard population in the 2002 Report to the Nation.
( Audio – Length: 02:28 )
Text Transcript
First Lynn A.G. Ries, M.S. and then Brenda K. Edwards, Ph.D. (both from NCI’s Surveillance Research Program) discuss the age adjustment to a year 2000 standard population in the 2002 Report to the Nation.
Q: This year, for the first time, you’re adjusting to a year 2000 standard population. Why is that and why are you doing it now?
Ms. Ries: Our program began in 1973, and the 1970 population was the closest we could come for age adjustment, but now that we have data for 1999, it seems a little out of date to see what rates in 1999 looked like compared to a population in 1970. So starting with 1999, the federal government decided that it would be good for all the federal agencies to use a more contemporaneous standard, which is the 2000 age distribution. This allows us to look at the 1999 rates as though they occurred in a population of today.
Dr. Edwards: There are two main reasons. The first one is that many of the health agencies were using different standards for adjusting populations to standardize their reporting of various health measures. We needed to bring our reporting to where we’re all using the same standard because it was making it very difficult, if not impossible, to compare reports from one agency to another. So the first reason is that we would all be using the same standard and the reporting of health data would all be comparable. The second reason, as Lynn said, if you look across the various groups reporting health data, it was time to bring those age adjustment rates to one that reflected our contemporary health picture today rather than 30 to 60 years ago, which is what the other standards had been doing.
- First Lynn A.G. Ries, M.S. and then Brenda K. Edwards, Ph.D. (both from NCI’s Surveillance Research Program) discuss the adjustment of SEER and other NCI statistics to a year 2000 standard population in the 2002 Report to the Nation.
( Audio – Length: 02:30 )
Text Transcript
First Lynn A.G. Ries, M.S. and then Brenda K. Edwards, Ph.D. (both from NCI’s Surveillance Research Program) discuss the adjustment of SEER and other NCI statistics to a year 2000 standard population in the 2002 Report to the Nation.
Ms. Ries: All of our data are self-contained within the Report to the Nation, so that if you go back to 1973, it has been adjusted to the year 2000 standard and all of our data has been updated–it is self-contained and it is the one report to use. You would not want to take a figure, a particular number, from a previous report and compare it. However, in each report, all of the data are consistent and age-adjusted to the same standard. In the Cancer Statistics Review, we actually showed data from 1973 through 1999, and all of the data are age adjusted to the year 2000 so that you could get data points back to 1973 and you don’t have to go to previous volumes to find the data points.
Q: And last year, you did this same adjustment to the year 1970 standard population?
Dr. Edwards: Right, because that was our standard in effect at that time. For most people, the two important messages are that this report is complete, comprehensive, and should meet most everyone’s need for the latest data and includes data for 1999 and updates for any earlier reports. The second point is that it will complicate people’s lives to try and compare any estimate with previous reports, but this should not be necessary because all of the data are in this report and the latest CSR.
Ms. Ries: Again, because cancer rates tend to go up with age, when you age adjust to the year 2000, many of the actual rates will look like they go up if you compare rates age-adjusted to the year 1970 with those age-adjusted to the year 2000. The main point is that if you age adjust everything you’re looking at to the same standard, then you can actually compare those rates. You don’t want to compare rates age adjusted to one standard with those age adjusted to another standard — it will look like an artificial increase in the rates, so what you want to do is compare apples to apples rather than apples to oranges.
- Brenda K. Edwards, Ph.D. of NCI’s Surveillance Research Program discusses prevalence and survivorship data from the 2002 Report to the Nation.
( Audio – Length: 01:36 )
Text Transcript
Brenda K. Edwards, Ph.D. of NCI’s Surveillance Research Program discusses prevalence and survivorship data from the 2002 Report to the Nation.
Dr. Edwards: Our prevalence of 8.9 million cancer survivors reflects an estimate based on some new methods that have been developed for obtaining estimates on how many people are alive today who have been diagnosed with cancer in the past. It actually differs from published methods used in the past. The previous estimate was based on data from just one state with a registry that went back 50 or 60 years. This new method is actually somewhat conservative because of the way in which we’re counting cancers, particularly cancer for people who may have had more than one cancer, and how we take that information into account. It’s actually a much better method because it allows us to estimate prevalence by many different characteristics of the population — not only cancer type, but age of the individual and how long they’ve been alive with the cancer. It will give us a lot more information for characterizing our cancer survivors and in planning for health services for cancer survivors.
Photos/Stills
