Methods

Top Methods Results Discussion Author & Article Info References

The patient sample selected for this analysis included 796 consecutive female patients with primary operable stage II (tumor measuring < 5 cm) or stage III (tumor measuring > 5 cm, plus the presence of fixed or matted nodes, skin involvement, or both) axillary lymph node-positive breast cancer who consented to be treated using three consecutive adjuvant chemotherapy protocols at the M. D. Anderson Cancer Center between 1974 and 1982 [9]. The numbers of patients registered for these protocols each year and included in this analysis were 5, 43, 83, 98, 67, 77, 114, 122, and 126 for the years 1974 through 1982, respectively. All patients had histologically confirmed invasive breast cancer and no evidence of distant metastases after appropriate staging procedures. Local therapy consisted of total mastectomy with axillary dissection; some women also received postoperative radiation therapy [9]. Adjuvant chemotherapy consisted of a combination of fluorouracil, doxorubicin (Adriamycin, Adria Laboratories, Columbus, Ohio), and cyclophosphamide administered at 28-day intervals [10]. In the first two protocols, when a cumulative dose of 300 mg/m² of doxorubicin was achieved, therapy continued with fluorouracil, methotrexate, and cyclophosphamide until a 2-year treatment period was completed. The third protocol included vincristine, prednisone, and tamoxifen in addition to fluorouracil, doxorubicin, and cyclophosphamide, and the duration of therapy was reduced to 1 year. Adjuvant postoperative radiotherapy and nonspecific immunotherapy with bacille Calmette–Guérin were given to patients in some subgroups, but this did not appear to influence outcome [11]. Detailed descriptions of these trials, including comparisons of outcome after adjusting for common prognostic indicators, have been reported [9-12]. Because adjusted relapse and survival rates for the three trials were remarkably similar, we pooled data for the patient groups to evaluate the effect of obesity.

To supplement the standard computer data set maintained for patients included in adjuvant trials, medical records of all patients were reviewed to obtain additional information on height and weight at the start of adjuvant therapy, smoking history, pathologic stage of disease [13], date of first relapse, and date of last contact or death. Sixty-one patients (8%) were excluded from the analysis because information was not available on initial weight, leaving 735 patients in the study (182 treated in protocol 1; 217 in protocol 2; and 336 in protocol 3). Nine (1%) of these patients were lost to follow-up.

For each patient, ideal weight was assumed to be the upper limit of the range for a medium body frame for her height, as given by standard weight and height tables [14]. If the patient's weight exceeded the ideal weight, the excess was computed as a percentage of the ideal weight. Obesity was defined as weight more than 20% over ideal weight. No attempt was made to identify patients who weighed less than their ideal weight; all patients weighing less than the Table entry were coded as 0. Body size, as measured by the Quetelet index (weight in kilograms divided by height in meters, squared) was also calculated and considered for analysis. Patients were considered to be postmenopausal if their last menses occurred more than 1 year before the diagnosis of breast cancer; perimenopausal (1 to 5 years since last menses) and postmenopausal patients were combined for this analysis.

Disease-free and overall survival were measured beginning on the date adjuvant chemotherapy was initiated (date of registration or randomization). The follow-up times ranged from 2 to 17 years (4 patients were lost to follow-up before the fifth year and another 5 were lost after 7 years), and the median duration of follow-up for those patients still alive was 10.7 years. Of the 735 patients, 362 (49%) had disease recurrence and 298 (41%) died of breast cancer. An additional 51 (7%) women died of causes not related to breast cancer and, for the purposes of this study of prognostic factors, these patients were counted as withdrawals from study (that is, censored) at their deaths to compute disease-free intervals and survival.

Survival probabilities were estimated using the Kaplan-Meier [15] method, and differences in survival experience were evaluated by the log-rank test [16]. Regression modeling techniques [17] were used to evaluate the association of body size with risk for disease recurrence and death; proportional-hazards assumptions were verified by graphic methods. The following terms were included in all models to control for differences in prognosis attributable to factors of recognized importance: menopausal status (0 = premenopause, 1 = postmenopause), stage of disease (0 = stage II, 1 = stage III), and number of involved axillary lymph nodes (0 = 3; 1 = 4 to 10; 2 = > 10). Terms defining body size were evaluated by testing the significance of increase in maximized log-likelihood associated with the addition of a particular term to a model. The potential presence of interaction effects between obesity and other characteristics was tested by defining product terms for the respective factors in a regression model. The linear or quadratic nature of the association between body size and risk for disease recurrence was also evaluated using a proportional-hazards model. Chi-square tests were used to compare incidence of obesity according to other patient pretreatment characteristics. Significance levels were based on two-sided tests.

Results

Top Methods Results Discussion Author & Article Info References

Sixty-one percent of patients were within 10% of their ideal weight; 14% weighed 10% to 20% more than their ideal weight; 11% weighed 20% to 30% more than their ideal weight, and 13% weighed over 30% more than their ideal weight. The incidence of obesity, as defined for this analysis, is presented in Table 1 for patients grouped by menopausal status, stage of disease, tumor size, and number of involved nodes. Obesity was more common among postmenopausal patients, those with stage III tumors, and those with tumors larger than 5 cm in diameter. Thus, obese patients tended to be at higher risk for recurrence because of other poor prognostic features. Information on estrogen receptor status was not available for many patients in the early clinical trials, but among the patients whose status was known (approximately 50%), 22% with estrogen receptor-positive tumors were obese, compared with 19% of those with estrogen receptor-negative tumors.

Disease-free and Overall Survival

With disease recurrence detected in 49% of patients, the estimated 10-year disease-free survival rate for the entire patient group was 51% (SE = 2%). Table 2 summarizes disease-free survival rates for patients grouped according to important prognostic features and two measures of body size. These results confirmed the importance of menopausal status, stage of disease, tumor size, and number of involved nodes in determining disease outcome among patients receiving adjuvant chemotherapy. Although disease-free survival was marginally superior among patients with estrogen receptor-positive tumors, the difference according to estrogen receptor status did not approach statistical significance (P = 0.39). Statistically significant differences in disease-free survival were noted for patients classified according to body weight Figure 1, Table 2 and Quetelet index, with excessive weight or body mass generally associated with shorter disease-free intervals. Although results do not rule out a linear association between obesity and outcome, the similarity between disease-free survival curves for the most obese patients and those 21% to 30% overweight suggested a constant influence of weight more than 20% over ideal. The estimated hazard rates for treatment failure were highest between 12 and 48 months of follow-up, decreased rapidly during the subsequent 48 months, and remained stable thereafter. Most notably, after the second year there was a higher risk for disease recurrence among obese patients, which continued throughout the entire observation period.

View larger version (18K): [in this window] [in a new window]   Figure 1. Disease-free survival of all patients who received adjuvant chemotherapy analyzed according to presence or absence of obesity (>20% over ideal weight) when therapy was initiated. Numbers of patients at risk at various intervals of follow-up are shown at the bottom.

Of the 735 patients, 298 women died of breast cancer, and 60% were estimated to be alive 10 years after the initiation of adjuvant chemotherapy (SE = 1.9%). A similar analysis of obesity and overall survival results showed similar conclusions regarding the association of obesity with poor prognosis (Figure 2).

View larger version (17K): [in this window] [in a new window]   Figure 2. Overall survival of all patients who received adjuvant chemotherapy analyzed according to presence or absence of obesity (>20% over ideal weight) when therapy was initiated. Numbers of patients at risk at various intervals of follow-up are shown at the bottom.

A close but not absolute correlation was found between tumor size and stage. Ninety-seven percent of patients in the highest Quetelet index quintile and 24% of those in the second highest quintile were considered obese by the definitions used in this analysis, whereas all patients in the lowest three quintiles were considered not to be obese (correlation coefficient = 0.94).

Multivariate Evaluation of Obesity

Using multivariate techniques to adjust for differences in disease characteristics between obese and nonobese patients, we confirmed an independent association between obesity and disease-free survival (Table 3). Based on the proportional-hazards regression model (which included terms for menopausal status, stage of disease, number of involved nodes, and obesity), the relative risk for disease recurrence among obese patients was 1.33 (95% CI, 1.05 to 1.68; P = 0.02) compared with that of the nonobese population. The increased risk associated with obesity was not as great as that for stage III disease (relative risk, 1.51; reference group, stage II) or the presence of more than 10 involved nodes (relative risk, 2.51; reference group, 1 to 3 involved nodes). Body mass, measured using the Quetelet index, also correlated positively with risk for disease recurrence, after adjustment for other prognostic factors (P = 0.01), and a quadratic term did not significantly improve the fit of a proportional-hazards model. Multivariate analysis, using overall survival as an end point, showed a similar association between obesity and survival (Table 3).

To determine the extent of interaction effects between obesity and other prognostic factors in our patients, terms defining interaction between obesity and stage of disease, obesity and nodal status, and obesity and menopausal status were tested in a proportional-hazards model that also included terms for each of the individual factors. Only the term defining a stage and obesity interaction approached significance at the 0.05 level (P = 0.054). The direction of the effect was for a greater difference in disease-free survival by obesity status among patients with higher-stage disease than for those with lower-stage disease. Figure 3 shows disease-free survival plotted for obese and nonobese patients separately according to disease stage [stage II Figure 3, top] or stage III Figure 3, bottom)). Among patients with stage III disease, body weight clearly affected disease-free survival, with higher rates of recurrence among obese patients. A smaller difference in disease-free survival existed among patients with stage II disease. Additional model-fitting procedures were performed in which a term for tumor size was considered in addition to, and in place of, a term for stage; in both cases, the test for obesity remained statistically significant (P = 0.04 and 0.03, respectively).

View larger version (21K): [in this window] [in a new window]   Figure 3. Disease-free survival of patients with stage II or III breast cancer who received adjuvant chemotherapy according to the presence or absence of obesity (>20% over ideal weight) when therapy was initiated. Top. Stage II breast cancer. Bottom. Stage III breast cancer.

Discussion

Top Methods Results Discussion Author & Article Info References

The number of involved axillary lymph nodes, size of the primary tumor, and the administration of adjuvant systemic therapies are known dominant prognostic factors for disease-free survival and overall survival in patients with primary breast cancer. Histologic and nuclear grades also correlate with these two end points, but interobserver reproducibility has been reported as poor. Estrogen and progesterone receptor status have been variously described to correlate with disease-free survival, overall survival, or both, but their discriminant power is limited [18].

A definite association between obesity and both overall survival and risk for treatment failure after mastectomy and adjuvant chemotherapy was evident in our data. In other data sets, obesity was associated with larger tumor size but not with the number of involved lymph nodes [19, 20]. Excess weight was found to be an independent factor associated with shorter disease-free intervals and overall survival. The differences in disease-free survival and overall survival between obese and nonobese patients remained statistically significant after correcting for other pretreatment characteristics, as previously reported [4-8]. The increased mortality rates for men and women less than average weight, both smokers and nonsmokers, have been well documented [21]. Because we combined patients having normal and subnormal weights, our analysis may have underestimated the magnitude of the differences in outcome between obese and nonobese patients [22].

Several definitions of obesity have been used in the literature, and some differed from that used for this analysis [4-6, 23]. We based our definition of obesity on the 1983 Metropolitan Life Insurance Company's height and weight tables. These tables were derived from essentially healthy insured persons. The proposed weight ranges in these tables were those associated with the lowest mortality rates. Those weight ranges are not necessarily "ideal" or desirable. Furthermore, these tables have been criticized for several reasons: First, it is not known whether they are representative of the population at large, especially minorities; second, the 1983 tables include higher weights for each height and frame category than the 1942 [24] or 1959 [25] editions, thus including, in the opinion of some experts, a number of overweight individuals in the "normal" range. Furthermore, weight and height tables, in general, represent one measurement in time and do not reflect subsequent changes in weight, dietary habits, blood pressure, or other factors that may influence health and risk of death [26]. Other issues relate to the exclusion of persons with chronic diseases who may not apply (or are not accepted) for insurance and the relation of low weight with cigarette smoking and shorter survival, a fact not considered in these tables.

Other methods to assess obesity include measurements of triceps and subscapular skinfolds or body mass indexes (body surface area, Quetelet index). None of these methods to assess obesity is precise. Skinfold thicknesses cannot be obtained retrospectively, whereas most patients with cancer have their weight and height recorded several times before and during treatment. We chose a definition based on height and weight tables because those measurements were readily available. Recognizing the limitations of this method, we also included a measure of body mass index (the Quetelet index) in our analysis. The National Health and Nutrition Examination Surveys used similar methods and definitions of obesity and found that 26% of U.S. adults 20 to 75 years old are overweight [27, 28]. This Figure is strikingly similar to our finding that 24% of our patients were obese. The close correlation of obesity, as defined in our study and the Quetelet index, strengthens our findings about the prognostic implications of obesity.

Although using the 1983 tables may have biased our study by including a higher percentage of overweight patients in the nonobese group, the bias would have been as a conservative estimate; thus, the finding of a 30% increase in risk for recurrence or death is probably an underestimate. Also, other researchers [21, 22] using different definitions of obesity reached similar conclusions. The relative value of the various definitions of obesity must be established, and the limitations of each method must be considered in the interpretation of data from this and other studies of obesity. The analysis of outcome according to the Quetelet index confirmed the reports of Tartter and colleagues [6] and Eberlein and coworkers [23]. The results of these analyses clearly showed that those patients with small body mass had the best prognoses, although we observed a somewhat inferior disease-free survival for patients with a Quetelet index less than the 20th percentile, suggesting that the association between body weight and prognosis may not be linear, as others [22] suggested.

Although the correlation between obesity and shortened disease-free survival and overall survival has been described and confirmed by others [5, 7, 8, 29, 30], our experience shows that the adverse prognostic value of obesity persists, despite the use of effective adjuvant systemic therapy. A relative risk of 1.3 is rather small, and obesity must be regarded as a rather weak prognostic factor compared with stage and axillary lymph node involvement. Because the hazard rates observed in this analysis are similar in magnitude to the benefits of adjuvant chemotherapy or hormone therapy (20% to 30%), it is tempting to speculate that in obese patients this increase in the risk for recurrence might neutralize the expected benefits of adjuvant therapy. Whether this is related to an increased risk for recurrence related to obesity or to a decreased efficacy of adjuvant therapy in obese patients remains unknown. That obesity was more common in older and postmenopausal women and that adjuvant chemotherapy produces more modest benefits in patients older than 50 years (or the postmenopausal population) indirectly supports this hypothesis.

Significant weight gains have been reported in women with primary breast cancer [22], especially women receiving adjuvant chemotherapy [5, 29, 31, 32]. Some researchers propose that decreased activity and increased food intake cause the women to gain weight, whereas others suggest that corticosteroid effects are responsible for weight gain. Although the adverse prognostic association of baseline (or pretreatment) obesity is unquestioned, the association of weight gain during therapy with poor prognosis has not been consistently shown. Three reports [29, 32, 33] found no association between weight gain during therapy and outcome. However, Camoriano and associates [34] reported an increased rate of relapse in premenopausal patients who gained weight during the first year after surgery. If the association of obesity with poor outcome is viewed as a cause-effect relation, then additional weight gain during primary therapy would not be beneficial and might increase the risk for disease recurrence and death. Prospective evaluation of this hypothesis is necessary.

Several hormonal mechanisms associated with obesity might influence the risk for recurrence and death from breast cancer [20, 30, 35-37]. Excess estrogen production secondary to peripheral conversion of androstenedione to estrogen may result from excess fatty tissue [30]. Because the levels of sex-hormone-binding globulin in obese women tend to be low, more free estrogen is available to target tissues for biological interactions. However, no data support the hypothesis that free estrogen increases the risk for breast cancer recurrence, nor is there information to suggest that supranormal concentrations of estrogens facilitate the development of metastases. Low levels of progestin with normal levels of estrogen have been reported to elicit carcinogenic effects [30].

Contrary to other reports [31, 38], the difference in outcome between obese and nonobese patients was more pronounced in the group with higher-stage disease. Goodwin and colleagues [29], Goodwin and Boyd [31], and Hebert and associates [38] reported that the prognostic effect of body size appeared to be greatest in women with less advanced disease, but the explanation for the discrepancy between their analyses and our own is not readily apparent. However, the relatively small sample size in the other reports [7, 26, 32] and variations in the distribution of prognostic factors may have resulted in the differences described. In addition, the results of subset analysis are usually less trustworthy than results of main outcome analysis.

Obesity could also play an important role in the metabolism of the cytotoxic drugs used for adjuvant therapy. Powis and coworkers [39] reported a significant decrease in the total body clearance of cyclophosphamide and a significant increase in its elimination half-life in patients with increased body weight. This may reflect a decrease in the metabolism of cyclophosphamide by cytochrome p450. The metabolism of doxorubicin is also altered in obese patients. Rodvold and associates [40] found that the elimination time of the drug was increased and that clearance decreased in a group of obese patients with breast cancer. This effect would result in increased drug concentration; however, they observed neither increased myelosuppression nor cytotoxicity.

The appropriate dose of chemotherapeutic agents is usually tailored to each patient. Some oncologists use body weight to determine this, whereas others use body surface area. Furthermore, some use the patient's actual weight, and others use the ideal body weight for the patient's height and frame. These variations result in substantial differences in calculated dose or dose intensity. Specifically, using ideal body weight may underestimate tolerance, leading to a lower dose intensity for obese patients than for patients of normal weight. Our group used actual body weights to calculate doses of cytotoxic drugs. In addition, our dose-modification schedule [9-12] resulted in a fairly homogeneous dose intensity for most patients [41]. Therefore, we do not believe that differences in dose intensity between obese and nonobese patients were responsible for the differences in disease-free survival and overall survival observed in this analysis.

The poor prognostic influence of obesity cannot be definitely explained and is probably multifaceted in nature. Additional studies to explore the pathophysiologic basis of this correlation are urgently needed. The controversy of determining the "right dose" of cytotoxic agents for use in obese patients can be addressed with pharmacokinetically supported trials; conversely, the weight and habits of patients should be considered in the analysis of clinical data. Prospective studies designed to evaluate the effects of dietary interventions on disease recurrence and overall survival in women with breast cancer are also clearly indicated.