We evaluated the association of breast cancer survival with prediagnostic recreational physical activity based on frequency, intensity, duration and timing. We found that women with who reported being ever being physically active prior to diagnosis had a reduced risk of mortality from any cause. These findings were observed among women diagnosed with both pre- and postmenopausal breast cancer, and regardless of their hormone receptor status. The risk reduction in mortality was stronger – near 50% reduction – among those postmenopausal women who reported engaging in any pre-diagnostic RPA vs no RPA during their postmenopausal years, rather than their premenopausal years. There was also evidence that moderate intensity RPA, but not vigorous, improved both all-cause and breast cancer-specific survival. Additionally, postmenopausal women had greater survival with increased RPA during the postmenopausal years. We did not, however, observe any effect measure modification by BMI, although the observed reduced mortality was somewhat stronger in women with a normal BMI (<25).
Our results are consistent with the population studies that have been previously published on this issue to date [6
], even though different exposure assessment methods and study populations were utilized. For example, Abrahamson et al. [6
] also focused on self-reported, pre-diagnostic levels of moderate and vigorous RPA only among younger women, which were assessed at three different time periods (ages 12 to 13 years, age 20, and in the year prior to the breast cancer diagnosis), rather than the lifetime assessment we included here. Abrahamson et al. found that the 22% risk reduction in mortality among young women was limited to the most recent pre-diagnostic RPA, which is more attenuated, but still fairly consistent with our findings reported here of a 30-50% reduction associated with activity in the past 10 years among postmenopausal women. Thus, both of these studies suggest a recency effect – that more activity undertaken in the years leading up to a breast cancer diagnosis may exert stronger beneficial effects on mortality risk than activity undertaken in the distant past. Alternatively, our results may indicate that recent prediagnostic RPA levels are strongly correlated with post-diagnostic RPA levels, which have been reported to also be associated with a modest reduction in the risk of death in two recent studies [19
]. Unfortunately, neither reported findings on whether women who were consistently active both prior to and subsequent to a breast cancer diagnosis had the best survival advantage, or whether women who reported being active after their breast cancer diagnosis were also those who were most active prior to their diagnosis. Another recent study examined change in physical activity from 1 year before to 3 years after diagnosis and found that those who decreased their physical activity more than 3 MET-h/wk had an increased risk of death compared to those who were inactive both before and after diagnosis [7
]. However, it was not reported whether the women who had decreased their physical activity were in the lowest or highest physical activity group before diagnosis. Thus, whether the beneficial effects on mortality associated with pre- and post-diagnostic levels of RPA actually reflect the same optimal time period, or two different time periods that are each associated with its own beneficial effects on survival remains unknown at this time. Similarly, it is also unclear whether breast cancer survival is affected by the change in the amount of RPA from before to after diagnosis. There are some studies that have shown that post-diagnosis physical activity levels briefly decline, but return to similar levels within three years [24
]. Future studies should focus on elucidating the effects of timing of the physical activity, and whether the RPA must have been initiated prior to the diagnosis of breast cancer to fully benefit from its positive effects on survival, or whether activity initiated after diagnosis is sufficient.
Although moderate activity levels were associated with a better survival advantage, associations for intense activity levels were weaker, an observation which is consistent with findings from animal studies [25
]. Our results for intensity of physical activity on survival after breast cancer are also similar to those found in two recent studies of prediagnostic RPA and breast cancer survival and another investigating post-diagnosis physical activity, both reporting a lower risk of death for moderate activity and while finding no association of vigorous-intensity activity [9
]. Similarly, several studies have found no decreased risk for breast cancer development with vigorous intensity of activities [27
]. However, a recent study found greater survival from breast cancer, but not all causes, with vigorous activity [9
] and the Nurse’s Health Study found that vigorous exercise after diagnosis also lowered risk of breast cancer death [19
]. We are unsure why vigorous intensity exercise did not improve survival in our study, however one possible explanation could be due to the low reporting of vigorous activities and lack of adequate variation of vigorous intensity activities in our population, where only one third of the population had ever engaged in vigorous RPA.
There are several limitations to this study that warrant mention. As with all studies that rely on self-reported measures of physical activity, measurement error, and the accompanying attenuation in the effect estimate, is of concern [30
]. We attempted to minimize the effects of poor measurement by relying on the instrument developed by Bernstein and colleagues [14
] which considers activity type, duration, intensity, and frequency. To optimize Bernstein’s method, we linked it to a life events calendar, a memory aid that is used to enhance recall of dates and activities. Given the limited number of outcome events, our observed risk reductions in mortality associated with RPA were strongest for all-cause, rather than breast cancer-specific mortality. Although the hazards ratios for breast cancer-specific mortality were consistently below the null, we could not rule out an effect size of 1, as was often observed for all-cause mortality. Additionally, in this study we had limited power to detect effect modification, therefore our stratified analyses should be interpreted as exploratory, and should be confirmed by other studies.
Many investigators have speculated on the biologic mechanism linking physical activity to the risk of developing breast cancer, including the insulin resistance pathway, inflammation, and DNA repair [5
]. Or, perhaps, the beneficial effects of RPA may simply be accomplished through avoiding weight gain [34
]. However, the precise mechanism remains elusive. For example, as we discussed above, controlling for prediagnostic body size did not substantially affect our effect estimates reported here. Similarly, previous studies have been unable to detect a strong and consistent association between RPA and levels of insulin-like-growth factor-1 [36
]. Thus, whether these same potential mechanisms also underlie the RPA-breast cancer survival link is currently unknown. A recent review, however, attempted to elucidate these mechanisms and summarized the literature for proposed common biomarkers in relation to breast cancer risk and physical activity separately. This review reported that after evaluating BMI, estrogens, androgens, sex hormone binding globulin, leptin, adiponectin, markers of insulin resistance, tumor necrosis factor-A, interleukin-6, and C-reactive protein, only BMI and estrone showed convincing evidence of an association with both breast cancer and physical activity [39
]. Future studies should focus on understanding the underlying biologic mechanisms, which would strengthen our evidence linking RPA to risk reductions in breast cancer incidence and mortality.
In conclusion, this population-based study of women with breast cancer provides some evidence for a modest beneficial effect of lifetime RPA on survival outcomes. The beneficial effect was even stronger for postmenopausal women who engaged in RPA during their postmenopausal years, rather than their premenopausal years suggesting that there is a particular period during adulthood when RPA could be most beneficial for prognosis. Whether this survival advantage is due to activity undertaken prior to diagnosis, as we and Abrahamson [6
] report, or whether it reflects activity undertaken after diagnosis, as reported by others [19
], or both, currently remains unclear. Nevertheless, RPA appears to be a promising strategy to reduce the risk of mortality among breast cancer survivors.