A meta-analysis of 79 randomized trials testing different methods for local control of breast cancer and involving 42,000 women shows that treatments that result in substantial reductions in five-year isolated recurrence rates lower long-term breast cancer mortality, even though the therapies do not affect cancer-related mortality during the first five years.
The data were gathered at the most recent meeting of the Early Breast Cancer Trialists' Collaborative Group (EBCTCG) in 2000 and unveiled by Sir Richard Peto, MSc, FRS, Professor in the Clinical Trial Service and Epidemiological Studies Units in the Nuffield Department of Clinical Medicine at Oxford University, in one of the plenary talks at the most recent San Antonio Breast Cancer Symposium.
The aim of the collaborative group, which was established in the 1980s, is to analyze and share data on individual women who participated in randomized trials of local or systemic therapy, and to use that information to create reliable meta-analyses of time to recurrence, breast cancer mortality, and other mortality during long-term follow-up.
“Breast cancer is a disease where we've really got to think of what we are achieving on the time scale of decades, not years,” Professor Peto said. “The question isn't what is the five-year survival, but rather 20-year survival.”
After First Five Years
The new data indicate clearly that the benefits of treatments from the 1980s come not in the first five years after randomization but in the time after the first five years, he said.
The differences in the 15-year mortality between control and active treatment groups is more than twice as great as the differences in mortality at five years. To look at this sort of long-term data, however, Professor Peto stressed, researchers must continue the world-wide collaboration and data sharing of EBCTCG.
In his first example, he described a comparison within the meta-analysis of 46 trials that tested radiation versus no radiation, but ignored other variables such as surgery type and nodal status, which should eventually be teased out.
Of the 24,000 women who participated in these trials, 12,000 have died during the 15-year follow-up.
In the control groups, 30.6% of women have had isolated local recurrence of the breast during the 15 years, compared with 10.3% in the radiation therapy arms of the study.
Thus, radiation provides a 20% absolute reduction of local recurrence. That 20% reduction in recurrence results in a 4% absolute reduction in breast cancer mortality, with 48.1% of women in the control arm dying of their disease, as compared with 44.0% in the radiation therapy arm.
Professor Peto noted that although the difference may seem small it is statistically significant because of the large sample size: “It is a small but real difference.”
Looking at the shape of the curves, one finds that most of the recurrence occurs early, within the first five years, and then both level off in both the control arm and the treatment arm. There is little difference in the mortality curves during the first five years, and only after that point do they start to diverge significantly.
Thus, even though there is no significant difference in mortality early on, the five-year local recurrence rate predicts whether the treatment will improve long-term survival.
“You can actually tell whether a treatment is working by looking at recurrence at five years,” he said. “The results show that local control does affect breast cancer mortality.”
Magnitude of Change in Recurrence Predicts Significance in Mortality Change
Furthermore, the researchers find that the magnitude of decrease in local recurrence rate is an important predictor of whether there will be a statistically significant reduction in long-term mortality.
Specifically, if the 79 trials are grouped by the magnitude of the reduction of risk of recurrence within the first five years after randomization—under 10%, 10 to 20%, or more than 20%—rather than by the type of treatment, there is a strong correlation evident with the effect on long-term mortality.
Specifically, when the researchers examined the outcomes for 17,000 women treated in trials that showed less than 10% difference between the control and treatment arm in terms of the rate of local recurrence at five years, there was no significant reduction in 15-year mortality with therapy.
However, the trials that resulted in more than 10% absolute difference in the rate of isolated recurrence between the active treatment and control arms at five years did show statistically significant improvement in breast cancer mortality at 15 years.
The rate of recurrence in these trials, which included 25,000 patients, was 33.0% in the control arms and 11.4% in the treatment arms. At 15 years, the rates of breast cancer mortality were 44.6% and 49.4%, respectively.
There is little heterogeneity between treatment types when analyzed this way, Professor Peto said. If a trial produces a substantial reduction in isolated recurrence risk within five years, then it will produce a moderate improvement in survival in the long-term.
Thus, better local control during the first five years has little effect on cancer mortality during that period, but leads to lower breast cancer mortality in subsequent decades.
Those data indicate that researchers need to measure the isolated recurrence rate and mortality during their trials and in long-term follows up studies.
Furthermore, there was a statistically significant increase in non-cancer mortality among patients treated with radiation, as compared with those in the control groups.
This increase in mortality included a rise in the number of deaths due to cardiovascular problems in the women on treatment. The increased rate in cardiovascular deaths did not account for the total increase in mortality risk for the radiotherapy-treated patients.
Given these findings, Professor Peto emphasized that researchers need to more closely track non-cancer mortality causes in patients over the long-term.
Specific Comparisons in Treatment Therapies
In addition to these broad class comparisons, the team analyzed 14 distinct types of local treatments, and for most of these there was a node-negative and node-positive comparison, resulting in a total of 24 different analyses.
For example, one of the categories was of the effects of the addition of radiotherapy after breast-conserving therapy and axillary clearance. A total of 7,311 women participated in these trials.
The 10-year isolated recurrence rate was 31.9% in controls who received surgery alone vs 10.5% in patients who received surgery plus radiation therapy.
That 20% absolute reduction in risk of recurrence translated into a 4% difference in mortality in the 10-year follow-up, with 24.7% of women in the control arm dying of breast cancer vs 20.9% in the treatment arm during that time.
This was highly statistically significant despite the fact that there was almost no difference in mortality during the first five years after randomization, he noted.
When the researchers split these patients into node-positive and -negative groups, they found that although both groups benefited, women with node-positive disease derived more benefit from radiation than did those with clear nodes.
A total of 6,097 women with node-negative disease were treated in trials. The local recurrence rates were 29.2% and 10%, for the control and treatment groups, respectively, and the breast cancer mortality rates were 20.3% and 17.4%.
Meanwhile, of the 1,214 node-positive women enrolled, 46.5% had local recurrence in the control arm and 33.5% did in the treatment arm. The breast cancer mortality for the two groups at 10 years was 45.2% vs 36.5%.
Even though it is too early to have 15-year follow-up on these trials, it is clear that radiation lowers breast cancer mortality in the decade after treatment in this patient group, Professor Peto said.
“If you have breast-conserving surgery and don't have radiation therapy, even if you have node-negative disease, it is quite a large risk of local recurrence.”
Treatments from 1990s Will Start to Be Reflected Soon
The treatments from the 1980s are affecting the overall mortality rate due to breast cancer, Professor Peto said. Treatment in the 1990s will start to be reflected in the coming years.
“By 2010, breast cancer mortality will be half of what it was in 1980.”
And while many women will die from their disease, he said, new treatments have prevented at least half of the deaths that would have occurred. “When we started this [collaboration] in the 1980s, nobody thought it was going to be this good. But it is.”
Meta-analyses Important for Generating Hypotheses
Meta-analyses, like the ones generated by Professor Peto and his collaborators, are important for generating hypotheses, commented Larry Norton, MD, Deputy Physician-in-Chief for Breast Cancer Programs and the Norna S. Sarofim Chair in Clinical Oncology at Memorial Sloan-Kettering Cancer Center.
“We often move very quickly in our field, and we need to look back at long-term follow-ups in the first, second, and third five-year period.”
Dr. Norton also pointed out that although meta-analyses and long-term follow up is critical for understanding the full value and implications of treatment, the structure of cancer research doesn't encourage such efforts.
The system encourages work that highlights efforts of single research groups and appears to be on the cutting edge of new science. That pressure makes the sort of analyses led by Professor Peto all the more important because they buck that trend, Dr. Norton said.
“One of the real values of the process is getting the investigators talking to each other.” That happened as a result of the EBCTCG effort to study the long-term effects of systemic therapies and Dr. Norton said he expects the effects will be similar in this situation.