PLEASE NOTE: Inaccuracies in my interpretation of the statistics in study referred to in this post were brought to my attention by one of the commenters below. After doing a lot of digging and head-scratching, I have decided to leave the original text, but have added strike-throughs over data that I wrongly interpreted, and further comments in purple text. The result is data that are more difficult to understand. The figures and graphs I referred to in the study show calculations for recurrence rates that are not simple rates of recurrence by incidence, but rather represent cumulative recurrence occurring over time. The percentages I quoted originally cannot be read, therefore, as the simple rate of recurrence per number of women. Rather, they represent the rates of recurrence per woman-year. As nearly as I can explain that, it means the percentage rate of recurrence per woman across a year. The rates that are shown in the graphs and figures in the study text plot the cumulative rate of recurrence, not the number of incidences of recurrence. And if that makes you wonder how it all applies to you as an individual, you are not alone.
Sometimes, I honestly feel as though I’ve unwittingly taken on the role of the unofficial Wet Blanket of Breast Cancer Research Statistics. I’m sure my college undergrad stats professor would be proud of me, however, for not taking statistical hype at face value. She was for many years a statistical analyst for the Framingham Heart Study, so she knew a thing or three about statistical hype.
This time, some of the hype came with a personal connection. The site for ABC News published a story with this headline: Breast Cancer: Tamoxifen Saves Lives but Some Women Go Without It. This particular story featured a photo of and interview with someone I know and respect, Alicia Staley, an activist, blogger, breast cancer survivor, and founder of The Staley Foundation. So, imagine how I felt doing what I always do when I see a news headline about breast cancer research. Which is that I become immediately suspicious of the percentages quoted in the article, and attempt to find the original study.
I’d like to say somewhat tangentially that I want to thank the publishers of The Lancet very, very much for allowing me to register as a non-paying but interested stakeholder, which allows me to have access to their published studies in PDF form. Not every research journal does this, although quite a few of them do. And for we cash-poor cancer survivors and bloggers, who usually do not make a red cent from our efforts, it is a boon not to have to pay an average of $31.50 for the privilege of viewing a complete study online for 24 hours. The research reported by ABC News was published yesterday by The Lancet at this link: The Efficacy of Tamoxifen. You can view the summary without being registered. However, if you want to get behind the headlines, you have to read the entire study, which you can see here: Longterm Tamoxifen Efficacy, An Analysis of Randomised Trials. The study was funded by Cancer Research UK, the British Heart Foundation, and the Medical Research Council, and was undertaken by the Early Breast Cancer Trialists’ Collaborative Group (EBCTCG), a collective of researchers whose stated mission has for many years been to review research literature about early breast cancer and its treatment. In the previous review linked to their name above, the EBCTCG describes their mission as follows:
Systematic reviews of the effects of health care attempt to bring together all relevant evidence on a particular intervention or treatment, so that people choosing between different interventions are more informed and can make better decisions.
In other words, my kind of people.
The ABC News story reported these findings from the study:
Taking a [tamoxifen] pill a day for five years reduced annual breast-cancer mortality by 30 percent for 15 years, the study found. And recurrence rates fell 47 percent in the first four years after treatment, and 32 percent over the next five years, according to the study.
You’re supposed to say, “Wow!” and be impressed. And it is impressive. And we’ve basically known this for quite some time. However…
Behind the hype: relative versus absolute.
What the study in fact did was to review some twenty-plus previous studies of tamoxifen efficacy, involving some 54,000 women, divided generally into the usual groups — those who took tamoxifen and those in the control groups who did not — and analyzed all the statistical findings. Both node-negative and node-positive women were included in each group, meaning that study participants were included whether they had breast cancer in their lymph nodes or not. It also included women who did and did not receive chemotherapy. This indicates indirectly that women who had early, small-tumor invasive breast cancers were included, along with those who had larger or more advanced tumors. Trials for women with ductal carcinoma were excluded. The data was then analyzed for three time periods: from 0-4 years after initial diagnosis; from 5-9 years after diagnosis; and for 10 or more years after diagnosis. Finally, overall statistics were computed for all the participants in the 20+ studies reviewed.
It has been known for some years now that the women who are likely to benefit the most from tamoxifen, as well as other adjuvant oral chemotherapy such as aromatase inhibitors, are those whose breast cancer tumors contain receptors that are positive for estrogen (ER+). This means these are tumors that in effect feed on estrogen, among other things. So, keeping estrogen away from breast tissue would, it would seem logical, help prevent tumors from recurring or would at least slow them down. That is what tamoxifen does, by blocking the receptors in breast tissue that take up estrogen.
The study included participants who had every type of tumor receptivity, and separated participants into groups according to the specific hormone receptivity of their tumors. If we look more closely at one of these groups, the ER/PR+ group, meaning women whose tumors are receptive to both estrogen and progesterone, we find a detailed analysis. For the first four years after initial breast cancer diagnosis, the data yielded the following:
- Of the
16,701women taking tamoxifen, there were 570 recurrences in this time period, or a recurrence rate of 3.41% per woman-year. These numbers are listed under a recurrence graph in Figure 1, for women with ER+/PR+tumors, but represent a calculation of percentage by woman-year, which is not a simple rate of recurrence per number of women, as they appeared to be when I first read them. The recurrence rates shown on the graph are calculated as the cumulate rate of recurrence over the number of years shown along the bottom of the graph, based on the total number of women in this group, indicated at the top as 7,378.
- Of the
15,432women who did not take it, there were 926 recurrences per in this period, or a recurrence rate of 6.0% per woman-year. See the previous comment.
- The difference between these rates of recurrence is 2.59%. This calculation, thank heaven, does not change.
- This means that there were relatively 43.1% fewer recurrences in the ER/PR+ group taking tamoxifen in the first four years after initial breast cancer diagnosis and treatment than there were in the group who did not take it. Still relative, still the same.
Out of the entire group of 32,133 women,there were 1,496 cumulative breast cancer recurrences in the first four years, for a cumulative recurrence rate within years 0-4 of about 20.3%.
- If the control group
of 15,432 womenhad instead taken tamoxifen, and responded as well as the tamoxifen group did, then it can be estimated that only 526 women in the control group would have had a recurrence, preventing some 400 women,400 recurrences, or another 2.59% of them, from having a recurrencemay have been prevented in the first four years.
It’s important to remember that these are statistics for breast cancer recurrence, not mortality. It’s also important to bear in mind how the word ‘recurrence’ is used in this study. Usually, a breast cancer recurrence is a return of symptoms in the same breast in which the original tumor appeared. A new occurrence of breast cancer in the other breast is not considered a recurrence, but a new event, and it may not be the same kind of breast cancer that was originally diagnosed in the first breast. In this study, however, first recurrence data included the occurrence of any breast cancer, whether is was local, in the opposite breast or distant.
Looking at the same group of ER/PR+ women, the statistics for years 5-9 after initial diagnosis are as follows:
- Of the
12,248women on tamoxifen for these years, there were 303 recurrences during this period, for a recurrence rate of 2.47% per woman-year. Same deal as the previous section.
- Of the
10,295women not on tamoxifen for this period, there were 360 recurrences, for a rate of 3.5% per woman-year.
- The difference between these two rates of recurrence was 1.03% per woman-year, which means that there were 29.4% fewer recurrences in the group that took tamoxifen compared to the group that did not, for years 5-9.
- If the control group had taken tamoxifen and experienced a similar decrease in recurrence rate, then
only 254 of these women would have had a recurrence, which would have meant106 fewer recurrences in these 10,295 women,may have occurred, or a recurrence rate that was 1.03% better per year, as it was in the group that took tamoxifen.
Similar data were calculated for all women with ER positive breast cancer, whether PR+ or not, including 44% who had node positive breast cancer, 56% who did not, and 51% who received chemotherapy.
- For all ER+ women,
there werein years 0-4, a recurrence rate of 3.74% per women-year occurred who had recurrencesamong those who took tamoxifen, compared to a rate of 6.71% in the group that did not. Which means that there were 2.97% fewer recurrences per woman-year in the group that took tamoxifen, for a relative decrease of 44.2% in rate of recurrence with tamoxifen.
- For all ER+ women, in years 5-9, a recurrence rate of 2.62%
had recurrencesper woman-year occurred among those whiletaking tamoxifen compared to 3.46% per women-year among those who did not, a decrease of 0.84% in the number of recurrences with tamoxifen, and a relative decrease of 24.3% in the rate of recurrence.
Below is the graph from Figure 5 of the study, showing the 100% ER+ group from which these figures above were taken. The text at the top about recurrence rate calculation is mine. The colored lines were added by me to mark off the time periods that were used in the study to generate the recurrence rates per woman-year shown below the graph. For about fifteen minutes, I actually understood how to calculate cumulative recurrence risks myself, and the numbers outlined in aqua are the cumulative rates at the end of Years 0-4, which, thank heaven, match the plot points on the graph. I just couldn’t face doing it again for Years 5-9, so I used estimates, taken from the plot points intersecting the green lines, in my edited numbers above.
At this point, it should be noted that among all women who had ER+ tumors, by year 4 there was a cumulative recurrence rate of 14.1% among women taking tamoxifen, which means that 85.9%
96.3% of women taking tamoxifen in years 0-4 did not have recurrences, and neither did 75.7% 93.3% of the women who did not take it. In years 5-9, about 78% 97.4% of women taking tamoxifen did not have recurrences, and neither did about 62% 96.5% of the women who did not. Something to keep in mind when an oncologist is flogging you with statistics, a practice I call ‘statistical terrorism.’ If you are one of those women among the single-digit percentages who are unlucky enough to have a recurrence, it certainly matters greatly that you are not among that lucky 90+ percentage of women who don’t. But it’s frightening enough to have breast cancer without having relative statistics thrown at us that are taken out of context. Yes, I know. To my regular readers, I readily admit that I sound like a broken record. Or a broken mp3 audio file.
But does it save lives?
Now to examine the data for breast cancer mortality. Once again, to zero in on those one would expect to benefit most from tamoxifen, I looked at the analysis of mortality done among all women with strongly ER+ tumors. Breast cancer mortality was calculated by starting with the total death rate among all of the study participants, and subtracting the death rate for those women who never had a recurrence of breast cancer, who presumably died of other causes. All of these rates and percentages are technically projected, based on taking tamoxifen for five years, but the projections are based on the hard data of comparing mortality rates with recurrence rates.
- In years 0-4 after initial diagnosis of breast cancer, 1.79% of women per woman-year taking tamoxifen died of breast cancer, compared to 2.46% of women per woman-year who did not take tamoxifen and also died of breast cancer. This represents a difference of 0.67% in the yearly rate of breast cancer deaths per woman-year between the two groups, or a decrease in the breast cancer mortality rate of 27.2% associated with taking tamoxifen in Years 0-4
these years. The cumulative mortality rates at the beginning of Year 5 are plotted on the graph, and were calculated to be 8.6% among women who took tamoxifen and 11.9% in women who did not, which represents a 27.7% estimated decrease in mortality by the beginning of Year 5.
- In years 5-9, 2.25% of women per woman-year who took tamoxifen for five years died of breast cancer, compared to 3.23% of women per woman-year who did not take tamoxifen, for a yearly difference of 0.98% per woman-year in the number of breast cancer deaths between these two groups. The average yearly decrease in the cumulative rate of breast cancer mortality associated with five years of tamoxifen use was 30.3% in these years.
In years 10-14, 1.54% of women who took tamoxifen for five years died of breast cancer, compared to 2.28% among those who did not take it. There were 0.74% more deaths by breast cancer in the latter group, and theAt Year 15, the end of the study period, the cumulative decrease in the breast cancer mortality rate associated with five years of tamoxifen use was projected to be 32.5%about 28% for this group, for a cumulative gain in the number of lives saved of 9.2%. For women who were 15 years or more from their initial diagnosis, the mortality rate was 1.48% in the tamoxifen group, 1.89% in the control group, a difference of 0.41% in the number of breast cancer deaths, or a 21.7% relative decrease in mortality associated with five years of tamoxifen use.
The study also looked for differences in these rates of recurrence and mortality for women with ER+ tumors between women under age 45 at diagnosis, and women between 55 and 69 at diagnosis, and found comparable statistics. This analysis can be found in Figure 6. of the study, for those who care to look at the full study.
Will it save my life?
Once again, one needs to keep in mind that taking tamoxifen for five years after being diagnosed with ER+ breast cancer was projected to have saved the lives of less than 1% more women in each woman-year calculated across the period of fifteen years studied compared to those women who did not take it. But the cumulative gain at the end of 15 years was 9.2%. It’s still a small number, a measley number in fact, across a span of 15 years, but because of the astounding incidence of breast cancer, it still represents a lot of lives. But perhaps not enough of them. At this writing, I have not been able to obtain statistics for how many women this
less-than-1% percentage represents among all the women who die every year of breast cancer. Nor is there space in this post to discuss the side effects of tamoxifen, and the mortality rates for women who died of uterine cancer or blood clots that may have been caused by tamoxifen, but there is some analysis of the available data on this in the study.
My purpose here is to place things into some kind of perspective. Back to my friend Alicia Staley, who was interviewed as part of the ABC News report. Staley was diagnosed with ER+ breast cancer at age 33. She took tamoxifen for two-and-a-half years, according to the report, but the side effects were having such a profoundly negative impact on her quality of life that she had to stop taking it. Six months later, she had a recurrence. Would she have had one if she had continued taking it? Naturally, she went round and round about her decision to stop it. But as the above data indicate, some women who take tamoxifen have recurrences anyway. Staley concluded that perhaps the tamoxifen bought her some time, enabling her to make a better and more definitive decision about what treatment to have when she did have a recurrence, without having to question herself had she never taken tamoxifen at all.
It’s a tough choice, but to echo the mission of EBCTCG, it’s important to have complete information about any treatment before deciding whether it’s the right one for you. And you can’t get that from a glance at the average news headline. And in the current climate of sound-byte, quick-and-dirty journalism, it seems almost pointless to plead with those who get paid to report on research to do a more thorough job.
Now, if I could only get someone to pay me for doing it…
…especially after forcing my brain around the concept of cumulative risk. I’m sure I probably still got some of the interpretation wrong, but I wanted at least to make the attempt to clarify the data I quoted here, especially since I am aware that parts of this post have been cut and pasted hither and yon, and perhaps used in a manner I did not intend. While I am perhaps more in awe now of the efforts of the EBCTCG, I am also moved to make a plea to all researchers to report their findings in plain language. And I would like to thank the splendid individual in the Epidemiology Section of the New Mexico Department of Health who wrote an explanation of the difference between incidence rate and cumulative incidence that I could actually understand, albeit briefly. I certainly won’t be giving a lecture on the subject anytime soon.
Please click on the post title or the comment link below to post a response.