Most epidemiologic studies report a reduced risk of developing breast cancer associated with higher levels of recreational physical activity, but little is known regarding its effect on prognosis.
Most epidemiologic studies report a reduced risk of developing breast cancer associated with higher levels of recreational physical activity, but little is known regarding its effect on prognosis.
In this study, the authors investigated whether activity undertaken prior to diagnosis influenced breast cancer survival in a population-based cohort. A follow-up study was conducted among 1264 women ages 20 to 54 years who were diagnosed with invasive breast cancer between 1990 and 1992. Women in the study were interviewed within several months of diagnosis and were asked about their average frequency of moderate and vigorous activity at age 13 years, age 20 years, and during the year before diagnosis. With 8 to 10 years of follow-up, all-cause mortality status was determined by using the National Death Index (n = 290 deaths).
A modest reduction in the hazards ratio (HR) was observed for the highest quartile of activity in the year before diagnosis compared with the lowest quartile (stage-adjusted and income-adjusted HR, 0.78; 95% confidence interval [95% CI], 0.56–1.08). High activity was associated with a reduced HR among women who were overweight or obese at the time of diagnosis (HR, 0.70; 95% CI, 0.49–0.99) but not among ideal weight or underweight women (HR, 1.08; 95% CI, 0.77–1.52). A reduced HR was not evident for activity at age 13 years or 20 years or for average activity across the 3 periods studied.
The results of this study provided some suggestive evidence for a beneficial effect on survival of recreational physical activity undertaken in the year before diagnosis, particularly among women who are overweight or obese near the time of diagnosis. Cancer 2006. Published 2006 by the American Cancer Society.
Although many observational studies have reported a reduced risk of developing breast cancer related to recreational physical activity,1 little is known regarding the effects of activity on prognosis after diagnosis.2 It is plausible that physical activity may influence prognosis by the same mechanisms that are proposed to affect etiology.3 These include decreased lifetime estrogen exposure, enhanced immune function, lower body fat, and reduced insulin resistance.2, 4
A number of important clinical and pathologic factors have been identified that influence disease prognosis negatively, including increased tumor size and late stage at diagnosis, high histologic grade, involvement of axillary lymph nodes, negative estrogen receptor (ER) or progesterone receptor (PR) status, and the status of specific tumor markers, such as the p53 tumor-suppressor gene or the HER-2/neu oncogene.5, 6 Although clinical markers generally are considered the most important prognostic factors for women with breast cancer, it has been estimated that stage and grade explain only approximately 20% of the observed variation in survival after diagnosis.7 Thus, there is a strong need to identify other prognostic factors that may improve our ability to determine which women are at increased risk of death and to provide more opportunity for patients to enhance their own survival.
One of the few modifiable lifestyle factors to receive considerable attention with regard to breast cancer survival is obesity. The results from a recent, comprehensive review suggested that the majority of previous studies reported higher mortality among heavier women.8 In contrast, few studies have evaluated the impact of recreational physical activity on prognosis.9–12 In the current study, we investigated the relationship between prediagnosis recreational physical activity and survival using a population-based study of young women who were diagnosed with breast cancer.
The study included women ages 20 to 54 years who were diagnosed with invasive breast cancer between May 1990 and December 1992 and who resided in a 5-county region of New Jersey or metropolitan Atlanta and had participated in a previously conducted case–control study, which has been described elsewhere.13 For the parent case–control study, women with incident breast cancer (cases) who were identified through rapid ascertainment systems participated in an in-home interview on a variety of factors related to breast cancer in young women, such as contraceptive and reproductive history, family history of breast cancer, use of cigarettes and alcohol, earlier body size, physical activity, and diet. Trained and certified interviewers also obtained anthropometric measures, including waist and hip circumference, as described previously.14 Interviews were completed by 86% of eligible women in the case group at a median of 4.2 months after diagnosis. Medical records were abstracted for clinical and pathologic characteristics related to the diagnosis. All information pertaining to physical activity15 and to the covariates that were included in this study was obtained from the administered questionnaire interview or from medical records.
Of the 1283 women diagnosed with invasive breast cancer who were eligible for this follow-up study, vital status data were available for 1264 women (98.5%), and those women constituted the final sample. This study was approved by the institutional review board of each participating institution.
The assessment of recreational physical activity in the parent study has been described in detail.15 Briefly, the frequency of selected vigorous or moderate activities was evaluated for 3 different periods: ages 12 to 13 years, age 20 years, and the year before diagnosis. Because the duration of activity was not measured, kilocalories expended per unit of time could not be estimated. Instead, relative units of physical activity were calculated by assigning metabolic equivalent (MET) scores to vigorous (9 METs) and moderate (5 METs) activities; the scores were weighted by the frequency of the activity and then summed to yield relative units of activity per week for each period. The women also were asked whether they participated in sports that required them to keep their weight low, such as gymnastics or ballet, and how frequently they climbed at least 2 flights of stairs (1 flight = 10 steps) without stopping in the year before diagnosis.
Vital status and, if the woman was deceased, date and cause of death were obtained through the New Jersey State Cancer Registry (New Jersey patients) and the Surveillance, Epidemiology, and End Results (SEER) Program (Atlanta patients), which perform routine computerized linkages with the National Death Index. Breast cancer was attributed as the cause for 85% of the 290 deaths according to the death certificate. There was no appreciable difference in the results when breast cancer-specific mortality rather than all-cause mortality was used as the outcome; only the latter results are reported here.
Summary stage data (local, distant, regional) were available for all women. Through a separate follow-up study, we were able to acquire complete and detailed information on cancer treatment and American Joint Committee on Cancer (AJCC)16 staging information for the Atlanta women only (n = 824 patients).
Follow-up started on the date of diagnosis and ended either on the date of death or the end of the study if the woman was still alive. Results did not change appreciably when follow-up was started on the date of the interview (data not shown). Participants were followed for a median of 8.5 years (range, from 3 months to 9.8 years) until the study ended on January 1, 2000 at both sites. The Kaplan–Meier (product-limit) method was used to generate survival curves for a preliminary examination of these data.17 Log-rank tests were employed to assess heterogeneity in time to death by levels of physical activity.18 Estimates of the hazards ratio (HR) for all-cause mortality and corresponding 95% confidence intervals (95% CIs) were calculated by using Cox proportional hazards modeling.19 The trend test was used to asses the hazard of death across levels of physical activity.20
For all main exposures and potential covariates, the proportional hazards assumption was evaluated by checking graphs plotting the log(−log S[t]) function against time for diverging or crossing survival curves and testing the statistical significance of time-by-covariate interaction terms.21 No violation of this assumption was observed.
Relative units of physical activity for the 3 periods studied were evaluated separately as dichotomous variables (using the median as the cut-off point), as tertiles or quartiles, and as continuous variables. All formats created similar results; only the results that were produced by using quartiles are reported here. We also assessed whether a history of being physically active or inactive in adulthood would affect mortality. Physical activity at age 20 years and in the year before diagnosis were considered jointly by grouping women as having either low activity or high activity (using the median as the cut-off point) for the 2 age periods. Using the 25th percentile as the reference category for a comparison with the other 3 quartiles combined produced similar results; the results using the median values are reported here.
To assess whether any of the factors listed below modified the relation between physical activity and mortality, interaction terms were tested with the likelihood ratio test using a significance level of P = .10.22 We defined effect measure modifiers as variables in which interaction with the main exposures created joint effects, which departed from perfect multiplicity. For assessment of confounding, covariates were included in multivariate models if they were related to the exposure (physical activity) or outcome (mortality) in bivariate analyses. Using backward elimination, covariates were removed from multivariate models in order of highest P value. Covariates were deemed confounders and remained in final models if they produced changes in the estimates of effect by ≥10%.23
The factors that were considered for confounding and effect measure modification included the following: menopausal status, age at diagnosis, race (white, nonwhite); stage (SEER summary stage defined as local, regional, or distant), ER and PR status, study site, prior breast biopsy, family history of breast cancer (mother or sister), age at menarche, use of oral contraceptives, parity, age at first live birth, number of miscarriages, number of induced abortions, and lactation history. Also considered were household income (<$24,000, from $24,000 to $49,999, or >$50,000); educational level; marital status; alcohol intake in the year before diagnosis; cigarette smoking (never, former, or current); average daily total of calories, fat, fruits, and vegetables consumed in the year before diagnosis; body mass index (BMI) at age 20 years; BMI at 1 year before the interview; waist-to-hip ratio (WHR) at diagnosis; the presence of comorbidities at the time of the interview (high blood pressure, high cholesterol, thyroid disease, diabetes, gallbladder disease, colorectal polyps, or other cancers); months between diagnosis and interview; and initiation of chemotherapy or radiation therapy prior to the interview.
Income and disease stage proved to be the only important confounders when we used our 10% change-in-estimate criteria; all analyses controlled for these 2 factors. There also was no confounding or effect measure modification by race or age. There was no appreciable difference in our estimates we evaluated by menopausal status. Thus, we do not show the results stratified by this factor or with the postmenopausal women omitted.
To explore more comprehensively any possible confounding by stage or treatment, we conducted a separate analysis that was restricted to the Atlanta women, for whom more detailed information was available. We observed no confounding or modifying effects by treatment status (surgery, chemotherapy, radiation, or hormone therapy) or by the presence of comorbidities (diabetes, heart disease, etc.) in the relation between physical activity and survival. Adjusting results for AJCC stage (4 categories: Stages I, IIA, IIB, III, and IV) instead of summary stage did not change the estimates (data not shown).
The median age at diagnosis was 42 years in our study population. Twenty-five percent of the women (n = 314 patients) were nonwhite (Table 1), of whom 291 were African-Americans (93%). Most tumors were classified as a local stage disease (57%) or regional stage disease (40%); approximately 50% of women had a household yearly income >$50,000. Women with the lowest levels of physical activity in the year before diagnosis were more likely to be nonwhite, overweight, less educated or lower income, and less likely to have ER-positive tumors.
|Characteristic||Physical activity level: No. of patients (%)||P†|
|Age at diagnosis, y|
|<40||412 (32)||125 (34)||92 (31)||114 (35)||91 (29)||.27|
|40–44||447 (36)||141 (38)||105 (36)||102 (31)||99 (36)|
|45–54||405 (32)||102 (28)||97 (33)||109 (34)||97 (35)|
|White||950 (75)||250 (68)||214 (73)||266 (82)||220 (79)||<.0001|
|Nonwhite||314 (25)||118 (32)||80 (27)||59 (18)||57 (21)|
|Stage of disease|
|Local||721 (57)||192 (52)||174 (59)||195 (60)||160 (28)||.15|
|Regional/distant||541 (43)||175 (48)||120 (41)||129 (40)||117 (42)|
|Premenopausal||985 (78)||286 (78)||231 (78)||261 (80)||207 (75)||.47|
|Positive ER status||695 (62)||186 (55)||153 (57)||200 (68)||167 (67)||.0004|
|Positive PR status||670 (59)||180 (54)||159 (60)||180 (61)||151 (61)||.20|
|≤High school||351 (28)||134 (36)||64 (22)||76 (24)||77 (28)||<.0001|
|≥Some college||913 (72)||234 (64)||230 (78)||249 (77)||200 (72)|
|Household yearly income|
|<$24,000||257 (21)||101 (28)||61 (21)||41 (13)||54 (20)||.0004|
|$24,000–$49,999||323 (26)||90 (25)||81 (28)||86 (27)||66 (25)|
|≥$50,000||652 (53)||169 (47)||150 (51)||189 (60)||144 (55)|
|BMI ≥25 kg/m2||482 (38)||210 (58)||143 (49)||125 (38)||96 (35)||<.0001|
|Waist-to-hip ratio >0.80||624 (50)||213 (59)||147 (50)||143 (44)||121 (45)||.0004|
The proportions of women who were alive at 5 years after diagnosis and at the end of the study follow-up were 82.8% and 77.0%, respectively (data not shown). An unadjusted Kaplan–Meier survival graph that represents the proportion of women who survived across time stratified by quartile of physical activity in the year before diagnosis is shown in Figure 1. Women in the lowest quartile of activity had consistently lower survival throughout follow-up. The survival probability increased by quartile of activity, with both the 3rd and 4th quartiles experiencing 80% survival versus 72% for the 1st quartile (log-rank test; P = .02). For physical activity at age 13 years, age 20 years, and the average of the 3 periods, the 4 curves by quartile of physical activity did not differ meaningfully (data not shown).
The HRs and 95% CIs for overall mortality and 5-year mortality by level of physical activity for the 3 periods (adjusted for stage and income) are presented in Table 2. The results for mortality at 5 years after diagnosis and at the end of follow-up were similar, and the results for overall survival are highlighted here. Having higher levels of physical activity at age 13 years was associated with a trend toward increased mortality (highest quartile vs. lowest quartile: HR, 1.30; 95% CI, 0.94–1.81), whereas having the highest quartile of physical activity in the year before diagnosis was associated with a slight decrease in mortality (HR, 0.78; 95% CI, 0.56–1.08). The unadjusted HR comparing the highest quartile with the lowest quartile of physical activity in the year prior to diagnosis was 0.66 (95% CI, 0.48–0.92). Physical activity at age 20 years (highest quartile vs. lowest quartile: HR, 1.18; 95% CI, 0.86–1.63), and the average of the 3 periods (highest quartile vs. lowest quartile: HR, 1.16; 95% CI, 0.84–1.60) did not affect survival.
|Period and quartile of physical activity (Relative units)||Overall mortality||5-Year mortality|
|No. alive||No. dead (%)||Crude HR (95% CI)||Adjusted HR (95% CI)||No. alive||No. dead (%)||Adjusted HR (95% CI)|
|Age 13 y|
|Q1 (1.6–21.5)||255||67 (21)||Reference||Reference||274||48 (15)||Reference|
|Q2 (21.6–45.6)||232||60 (21)||1.06 (0.75–1.49)||1.03 (0.72–1.46)||248||44 (15)||1.00 (0.67–1.52)|
|Q3 (45.7–75.5)||246||80 (24)||1.26 (0.90–1.72)||1.17 (0.85–1.63)||265||61 (19)||1.20 (0.82–1.75)|
|Q4 (75.6–98.0)||212||78 (27)||1.41 (1.02–1.94)||1.30 (0.94–1.81)||231||59 (20)||1.34 (0.91–1.97)|
|P for linear trend||.02||.08||.09|
|Age 20 y|
|Q1 (1.6–3.4)||252||74 (23)||Reference||Reference||264||62 (19)||Reference|
|Q2 (3.5–14.0)||242||78 (24)||1.09 (0.80–1.49)||1.11 (0.81–1.52)||271||49 (15)||0.82 (0.56–1.19)|
|Q3 (14.1–35.0)||236||54 (19)||0.82 (0.58–1.16)||0.87 (0.61–1.24)||247||43 (15)||0.85 (0.57–1.25)|
|Q4 (35.1–98.0)||215||79 (27)||1.20 (0.88–1.65)||1.18 (0.86–1.63)||236||58 (20)||1.00 (0.70–1.42)|
|P for linear trend||.55||.55||.99|
|Year before diagnosis|
|Q1 (1.6–3.4)||259||100 (28)||Reference||Reference||281||78 (22)||Reference|
|Q2 (3.5–13.5)||226||66 (23)||0.75 (0.55–1.02)||0.86 (0.63–1.18)||244||48 (16)||0.81 (0.57–1.17)|
|Q3 (13.6–35.0)||250||65 (21)||0.65 (0.48–0.89)||0.81 (0.60–1.12)||271||44 (14)||0.72 (0.49–1.04)|
|Q4 (35.1–98.0)||210||54 (21)||0.66 (0.46–0.92)||0.78 (0.56–1.08)||222||42 (16)||0.79 (0.54–1.15)|
|P for linear trend||.0004||.10||.12|
|Average of the 3 periods|
|Q1 (1.6–16.6)||237||71 (23)||Reference||Reference||254||54 (17)||Reference|
|Q2 (16.7–29.4)||247||62 (20)||0.87 (0.62–1.22)||0.84 (0.60–1.18)||267||42 (14)||0.74 (0.50–1.11)|
|Q3 (29.5–43.0)||235||70 (23)||1.03 (0.75–0.43)||0.97 (0.70–1.35)||251||54 (18)||0.97 (0.66–1.41)|
|Q4 (43.1–98.0)||226||82 (27)||1.18 (0.86–1.61)||1.16 (0.84–1.60)||246||62 (20)||1.12 (0.78–1.62)|
|P for linear trend||.20||.25||.30|
|Age 20 y (year prior to diagnosis)|
|Low activity (low activity)||308||102 (25)||Reference||Reference||333||77 (19)||Reference|
|Low activity (high activity)||186||50 (21)||0.79 (0.56–1.10)||0.91 (0.65–1.28)||202||34 (14)||0.83 (0.56–1.25)|
|High activity (low activity)||177||64 (27)||1.10 (0.82–1.51)||1.09 (0.80–1.49)||192||49 (20)||1.10 (0.77–1.57)|
|High activity (high activity)||274||69 (20)||0.75 (0.56–1.02)||0.85 (0.63–1.16)||291||52 (15)||0.85 (0.60–1.21)|
Compared with women who had low physical activity both at age 20 years and during the year before diagnosis, we observed no meaningful difference in mortality for women who started with either low or high physical activity (using the median as the cut-off point) during the earlier period and changed categories by the later period. In addition, there was no association between all-cause mortality and participating in sports that required low body weight during childhood, such as gymnastics or ballet, nor was there a correlation with climbing stairs in the year before diagnosis (data not shown).
Results for the association between mortality and physical activity in the year prior to diagnosis were modified by BMI for the same period (Table 3). High physical activity was associated with reduced mortality among women who were overweight or obese (BMI, ≥25 kg/m2) at the time of diagnosis (HR, 0.70; 95% CI, 0.49–0.99), but there was no such relationship among ideal weight or underweight women (BMI, <25 kg/m2; HR, 1.08; 95% CI, 0.77–1.52). Very similar results were observed in these stratified analyses when the 48 underweight women (BMI, <18.5 kg/m2) were omitted. We observed a similar pattern when we evaluated the modifying effects of WHR rather than BMI; the same magnitude of risk reduction was observed among women with a higher distribution of abdominal fat (WHR > 0.80) but not among women with lower abdominal fat (WHR ≤ 0.80) at the time of the interview (data not shown).
|Covariate||Physical activity level|
|No. of patients*||HR||No. of patients*||HR (95% CI)|
|<25 kg/m2||347||1.0||401||1.08 (0.77–1.52)|
|≥25 kg/m2||299||1.0||174||0.70 (0.49–0.99)|
|P for interaction||.05|
In this population-based study of young women, recreational physical activity near the time of diagnosis was weakly associated with survival, whereas activity in early adolescence or at age 20 years was not. For the year before diagnosis, the adjusted HRs decreased by quartile of activity, suggesting a beneficial effect of exercise on breast cancer survival. However, these results should be interpreted cautiously, because the effect was modest, and the confidence interval included the null; thus, a chance finding cannot be ruled out. It is noteworthy, however, that high physical activity reduced mortality for overweight or obese women but not for women of ideal weight. Because obesity is relatively well established as a poor prognostic factor in breast cancer,24 it is promising physical activity may provide an opportunity to improve survival in this subpopulation.
Previous observational studies that measured recreational activity just before diagnosis in relation to mortality have been mixed.9–11 Although the assessment of activity was similar to the method we employed, 2 of those studies9, 10 used considerably smaller sample sizes and may not have had sufficient power to detect an association. A third study by Enger and Bernstein11 included a similar population of 717 premenopausal women, and those authors reported that higher levels of exercise in the year before diagnosis were associated weakly with reduced mortality. In another study by Holmes and colleagues,12 which was relatively large and evaluated the effect of exercise at least 2 years after diagnosis, a decrease in all-cause and breast cancer related mortality was reported with increasing levels of activity. Thus, our current results are in agreement with those from the 2 larger studies, in which it was reported that recreational physical activity may enhance survival, but it is unclear whether activity undertaken either immediately before diagnosis, or after diagnosis, or both is beneficial for influencing survival.
Compared with our study, in which we observed reduced mortality only among overweight/obese women, in the study by Holmes et al.,12 the only previous survival study in which results were reported by body size near diagnosis, a similar beneficial effect of physical activity was observed for both ideal weight women and overweight women. Studies on physical activity and breast cancer incidence have produced mixed results on the potential modifying effects of BMI. Several authors have reported lower risk associated with high physical activity among leaner women only,25–27 whereas others have reported no effect modification by body size.28–34
It is possible that there has been misclassification of physical activity exposure.15 Based on studies that focused on the relation between breast cancer incidence and physical activity,35 investigators have reported that it is critical to use a comprehensive assessment method of activity that obtains the frequency, duration, and intensity of all leisure, occupational, and household activities. The questionnaire that was used in our study did not ask about participation in an exhaustive list of activities, nor did it account for time spent engaged in each activity. In addition, it did not include occupational or household duties, which may have led to additional bias. For instance, women with the lowest amount of recreational activity may have engaged in high levels of household or job-related activities and may have differed with respect to education, income, access to health care, and social support.
To gain a better understanding of the extent to which potential misclassification of physical activity levels in the year before diagnosis may have biased our results, we conducted a small sensitivity analysis. Simple deterministic methods were used to produce “corrected” HRs under various hypothetical misclassification scenarios.36 We assumed varying degrees (80–90%) of specificity and sensitivity of dichotomous misclassification of physical activity (high vs. low) by vital status (alive or deceased) at the end of the study period.
We learned from this sensitivity analysis that each scenario of nondifferential misclassification showed our observed results to be biased toward the null. For example, assuming that 10% of all women who reported high versus low activity levels were classified incorrectly, the observed adjusted HR of 0.79 decreased further to 0.46 in the “corrected” models. The corrected HRs ranged from 0.30 to 0.46 under the assumption of nondifferential misclassification. We have no reason to suspect that reporting of physical activity would differ by vital status, because participants were interviewed near the time of diagnosis. However, when we made assumptions that were weakly differential by vital status, the corrected HRs were similarly low or were nearly the same as the observed HRs. In summary, the results from this sensitivity analysis suggested that the beneficial impact of physical activity on survival would be stronger than what we observed if nondifferential exposure misclassification was present in this study.
Early breast cancer symptoms during the year before diagnosis may alter behavior. Thus, estimated activity levels for that period may not represent average activity during adulthood. This may be especially true for women who are the sickest and, thus, are less likely to exercise and more likely to die. It is also possible that postdiagnosis behavior at the time of the interview may have influenced reporting of prior physical activity practices or that activity was over reported in an attempt to give a socially desirable response. These potential biases may have altered the observed HRs.
Although physical activity was captured for several prediagnosis periods, we did not assess any changes in activity levels after diagnosis. It is not clear when during the life span physical activity would have the greatest impact on developing or surviving cancer. If the period after diagnosis proves important for prognosis, then it is possible that the prediagnosis levels of activity measured in the current study are not reflective of postdiagnosis patterns. In a recent study, it was reported that physical activity levels decreased among patients with newly diagnosed breast cancer, probably because of treatment effects.37 Similarly, 2 studies of breast cancer survivors showed that most women were inactive during the treatment phase; however, for most women, the levels of activity after treatment were similar to their prediagnosis levels.38, 39 Thus, our prediagnostic level indeed may be a good proxy for postdiagnostic levels.
There is additional potential for unmeasured or poorly measured confounders or effect measure modifiers to bias the results of this study. In particular, we were unable to control for additional socioeconomic features, such as access to quality health care or psychosocial factors (depression, support, etc.), that may affect cancer survival. Given the relatively young age of these women, we would not expect comorbidities to be a strong source of residual confounding. The presence of common comorbidities (high blood pressure, diabetes, high cholesterol, etc.) either at baseline or during follow-up (in the analysis restricted to the Atlanta women) did not confound these results.
Because 14% of the women who were eligible to participate in the parent case–control study opted not to enroll, the current follow-up cohort represented only 86% of the women who were diagnosed with breast cancer who could have participated. Bias may have been introduced into this study if the relation between physical activity and survival in the nonresponders differed from that of the women who chose to participate. It is likely that many of the women who did not participate in the original parent study may have been the sickest and, thus, more likely to die during follow-up and less active in the year prior to diagnosis. If this is true, then we expect that the results reported here underestimate the beneficial effects of exercise.
This study has several strengths. First, loss to follow-up with regard to vital status was <2%. In addition, the women were selected from a population-based sample and were followed for a relatively long period. We believe there was a high degree of accuracy with respect to the outcome, which was all-cause mortality outcome. We also were able to assess a comprehensive list of potential confounders and effect modifiers. Although the tool for measuring physical activity did not include a comprehensive assessment of lifetime physical activity, this study had the advantage of assessing the impact of activity for different periods during a woman's life.
In summary, the current study provides some suggestive evidence for a modest beneficial effect of recreational physical activity undertaken in the year before diagnosis on breast cancer survival among young women, particularly those who are overweight at diagnosis. The results are encouraging, because few modifiable lifestyle factors for improving prognosis have been identified. Although survival rates for breast cancer generally are improving because of improved techniques for early detection and better treatment modalities, there is considerable interest among survivors to know what they can do to alter the course of their disease.
Most research to date that has addressed physical activity among cancer survivors has focused on exercise as a means to improve physiologic effects or on quality-of-life issues during and after treatment40 rather than on exploring the actual effect of exercise on long-term prognosis. Further studies on the topic of physical activity and breast cancer survival are warranted given the plausibility of the association. Future investigations should include more thorough and comprehensive assessments of lifetime physical activity that incorporate measures of intensity, duration, frequency, and timing of the activity both before and after diagnosis to aid in determining the true relationship between physical activity and breast cancer prognosis.
If further research confirms that physical activity reduces mortality among women with breast cancer, then programs and policies to promote such activity for this purpose may be adopted. Research has demonstrated that programs and policies successfully may increase physical activity in selected communities, including cancer survivors.41