Menopausal Status and the Impact of Early Recurrence on Breast Cancer Survival

Gary H. Lyman, MD, MPH, Nicole M. Kuderer, Stephen L. Lyman, MSPH, Markus Debus, Susan Minton, DO, Lodovico Balducci, MD, John Horton, MB, ChB, Douglas Reintgen, MD, and Charles Cox, MD

Early recurrence has a different significance in premenopausal and postmenopausal women with breast cancer.

Background: Breast cancer represents the leading form of invasive cancer among American women, killing nearly 50,000 annually. Several prognostic factors that are associated with survival include age, race, menopausal status, and the stage of disease at presentation.
Methods: Patient characteristics were collected based on a systematic chart audit of demographic features and medical, family, and social histories. We studied the survival of 220 patients with recurrent disease out of 1,429 consecutive patients with breast cancer seen over a 15-year period.
Results: Patients with a disease-free interval following diagnosis of less than 24 months were more frequently premenopausal and hormone receptor-negative than those with a disease-free interval of 24 months or greater. Patients with early recurrence had a shorter survival than patients with late recurrence. Menopausal status, nodal involvement, receptor status, and the site of recurrent disease were independent predictors of survival following recurrence.
Conclusions: Premenopausal women with early recurrence of breast cancer experience a significantly shorter survival than those with late recurrence, even after adjustment for hormone receptor status and site of recurrence. This effect was not seen in postmenopausal women.

Introduction

More than 180,000 women will be diagnosed with breast cancer in the United States during 1997.¹ Although breast cancer survival in this country has increased over the past three decades to more than 80% at five years, approximately 46,000 women will die of breast cancer this year.² Several demographic factors, including age and race, have shown prognostic importance for breast cancer recurrence and survival.^3,4 Clinical factors of recognized prognostic importance for breast cancer recurrence and survival include lymph node involvement, tumor size, hormone receptor positivity, and several histologic factors including tumor grade.

The majority of patients who develop recurrent disease will eventually die of metastatic breast cancer, but the time from recurrence to death varies greatly. Survival subsequent to disease recurrence appears to correlate with the site and extent of recurrence and to factors related to the biologic activity of the disease including hormone receptor status. Our study assesses the impact of the disease-free interval on subsequent survival in women with recurrent breast cancer. Early recurrence appears to have adverse prognostic importance for premenopausal but not postmenopausal women with breast cancer.

Methods

We studied the survival and recurrence patterns for 1,429 consecutive women with early-stage invasive breast cancer who were seen and evaluated at the University of South Florida and the H. Lee Moffitt Cancer Center & Research Institute between 1980 and 1995. Patients whose disease recurred prior to referral to this Center were excluded from our analysis. All diagnoses made at outside institutions were confirmed by internal review of the original biopsy material. Individual patient characteristics (eg, demographic features, medical history, family history, social history) were collected based on a systematic chart audit. Diagnostic and staging information and initial treatment were also recorded. Background information included age, race, sex, family history of breast cancer, history of smoking, and menstrual history (menopausal status, number of pregnancies, history of oophorectomy, and history of exogenous hormones). Diagnostic and staging information consisted of stage, number of nodes, number of positive nodes, size of primary tumor, presence of bilateral disease, presence of positive margins, histologic type, tumor grade, DNA content, S-phase fraction, HER-2/neu oncogene expression, epidermal growth factor receptor, cathepsin D, and levels of estrogen receptor (ER) and progesterone receptor (PR). Treatment information included dates, type, and place of surgery, adjuvant radiotherapy, adjuvant chemotherapy, and adjuvant tamoxifen use. Follow-up information consisted of date of diagnosis, date and site of first demonstrated recurrence, date and cause of death, and date of last contact. Contralateral breast cancer was considered to be a second primary and not disease recurrence. Ipsilateral breast or chest wall cancer was considered to be recurrent disease. The disease-free interval was evaluated by empirically dividing patients with recurrent disease into those with early recurrence (<24 months) and those with late recurrence (greater than or equal to 24 months).

The distribution of each variable was studied, and measures of central tendency and variance were estimated. Except as otherwise indicated, ER and PR were evaluated as dichotomous variables (negative/positive). Summary measures were calculated for each variable consisting of the mean or median for continuous measures and proportions for categorical measures. Except where otherwise specified, the measure of variance used for means throughout this study is the standard error of the mean. Mean subject ages were compared using an independent sample Student's ttest incorporating a pooled variance. Other measured data were found to deviate substantially from normalcy and were studied by nonparametric methods. The Mann-Whitney U test (Wilcoxon rank sum) was used to test the equality of distributions among groups.⁵ Proportions observed among categorical variables were compared using the chi-square test or Fisher's exact test. Confidence intervals on proportions were calculated by the method of Cornfield.⁶ Ninety-five percent (95%) confidence intervals were used throughout.

Survival functions were estimated for overall survival, disease-free survival, and survival following recurrence using the product-limit method of Kaplan and Meier.⁷ Overall survival was calculated from the date of diagnosis to the date of death. Disease-free survival was calculated from the date of initial diagnosis to the date of recurrence or death, whichever came first. Survival after recurrence was calculated from the date of documented recurrence to the date of death. Patients not experiencing an event were considered censored at the date of last contact. Inference on survival functions among subgroups was based on the log-rank test (Mantel-Haenszel) for the equality of the survival functions.^8,9

Multivariate analysis was performed by fitting survival data to the proportional hazards regression model of Cox.¹⁰ In such models, the covariate function is proportional to the hazard or mortality and, therefore, positive coefficients indicate a shorter survival with increasing value of the covariate. After satisfying the proportionality assumption of the model, potential interactions between disease-free interval and each major covariate were studied. Interaction terms failing to achieve statistical significance were not considered further in the models. Variable entry into the models proceeded in a forward stepwise fashion using an adjusted chi-square statistic in variable selection.¹¹ The likelihood ratio test was used to test the hypothesis that the covariates in the model have no influence on the survival outcome. Adjusted estimates of the relative risk were derived from the multivariate proportional hazards models using the standard error of the coefficient to estimate confidence.

Results

Overall and disease-free survival of the 1,429 patients with invasive breast cancer is shown in Fig 1. The median survival of all patients was 142 months, and the median disease-free survival was 71 months. A total of 220 patients (15.4%) developed recurrent disease. Sites of first recurrence included bone (29.0%), ipsilateral breast/chest wall (19.0%), lung (18.0%), lymph nodes (13.0%), and other sites (20%, including three patients with hepatic metastases as the site of first recurrence). At the time of this analysis, 75 patients (34%) with re-current disease had subsequently died. Breast cancer was the documented cause of death in 68 patients (91%), and the cause of death in the remaining seven patients is unknown.

Of 216 patients with documented dates of recurrence, 85 (39.4%) recurred within 24 months of diagnosis, and 131 (60.6%) recurred more than 24 months following diagnosis. As shown in Table 1, the distribution of several prognostic factors differed among those with early recurrence compared to those with late recurrence. The median tumor size at diagnosis was 2.7 cm in patients with early recurrence and 2.4 cm in those with late recurrence (P=.058). The median number of axillary lymph nodes involved at diagnosis was one in those with early recurrence and zero in those with late recurrence (P=.124). The median ER level was 22.0 fmol/mg cytosol protein in those with early recurrence compared to 40.0 in those with late recurrence (P=.145). The median PR level was 41.0 fmol/mg cytosol protein in those with early recurrence compared to 50.5 in those with late recurrence (P=.749). There was no significant difference in the proportion of patients who received adjunctive chemotherapy or tamoxifen between those with early vs late recurrence. Likewise, there was no significant difference in the proportion of either premenopausal or postmenopausal women who received adjunctive chemotherapy or tamoxifen after adjusting for hormone receptor status.

As shown in Fig 2, survival subsequent to developing recurrent disease was shorter in patients with early recurrences (median = 25 months) than with late recurrences (median = 51 months) (P=.0277). Survival differences were also seen for other prognostic factors as displayed in Table 2. The median survival subsequent to recurrence was 16.6 months among patients with negative estrogen receptors compared to 47.2 months in patients with positive estrogen receptors (P=.0002). Likewise, premenopausal patients demonstrated significantly shorter survival than postmenopausal women with median survivals of 25.3 and 51.1 months, respectively (P=.0119).

Table 3 displays a series of proportional hazards regression models including an indicator variable for early vs late recurrence. Regression models were fit in a forward stepwise fashion for menopausal status, lymph node involvement, estrogen receptor positivity, and site of recurrence. Early vs late recurrence remains a highly significant independent predictor of survival subsequent to recurrence after adjustment for the above significant prognostic factors. Table 4 provides estimates of the relative risk for mortality for early vs late recurrence with progressive adjustment for menopausal status, lymph node involvement, estrogen receptor positivity, and site of recurrence. After adjustment, early recurrence patients experience approximately a two-fold greater risk of mortality compared to late recurrence patients.

Possible interaction between disease-free interval and each of the other prognostic factors was examined. Significant effect modification was observed only for menopausal status. As shown in Fig 3A, shorter survival was associated with early recurrence (median = 19 months) compared with late recurrence (median = 31 months) among premenopausal women (P=.0097). No significant difference in survival was observed between early (median = 49 months) vs late (median = 51 months) recurrence in postmenopausal women (Fig 3B). The relative risk of mortality for early vs late recurrence estimated by separate regression models based on menopausal status was 0.95 (0.45, 2.01) for postmenopausal women and 2.29 (1.20, 4.37) for premenopausal women. However, after adjusting for lymph node involvement, receptor status, and site of recurrence, the relative risk of mortality for early vs late recurrence estimated by separate models was 0.73 (0.16, 3.25) for postmenopausal women and 8.49 (0.89, 81.39) for premenopausal patients.

Discussion

The risk of breast cancer recurrence is associated with a variety of prognostic factors: age, race, menopausal status, characteristics of the tumor at the time of diagnosis including size, lymph node involvement, receptor status, and histologic and biologic features of the disease. Survival subsequent to recurrence likewise appears to correlate with menopausal status, the stage and characteristics of the malignancy at the time of diagnosis, and the site and extent of recurrent disease.

In this report, the timing of breast cancer recurrence following initial diagnosis was studied. Patients with disease recurrence within 24 months of diagnosis experience a significantly shorter subsequent survival than patients whose disease recurs beyond 24 months. Patients with early vs late disease recurrence differ with regard to several parameters including menopausal status and hormone receptor status. Survival subsequent to recurrence was also found to be associated with several prognostic factors including receptor status and menopausal status. There was no significant difference in adjuvant treatment with chemotherapy or tamoxifen between patients with early vs late recurrence after consideration of menopausal status and hormone receptor status.

In multivariate analysis, it is evident that the disease-free interval remains a significant predictor of subsequent survival after adjustment for menopausal status, receptor status, lymph node involvement, and the site of metastatic disease. With these variables in the model, no other variable considered was a significant predictor of survival. In fact, with adjustment for the above prognostic factors, the relationship between disease-free interval and survival is even stronger, suggesting that this relationship was partially confounded by other factors. The relative risk for mortality for early vs late recurrence after adjustment is approximately 2, representing a doubling of the hazard rate in those with early recurrence of disease.

An interaction was observed between menopausal status and the timing of recurrence. The relationship between the disease-free interval and subsequent survival with breast cancer appears to be confined to premenopausal women. Among postmenopausal patients, survival is nearly identical in those with early and late recurrence. Among premenopausal women, those with an early recurrence have a much shorter subsequent survival than patients with late recurrence. The relative risk for mortality associated with early disease recurrence becomes even greater with adjustment for other prognostic factors.

The reasons for the interaction between the disease-free interval and menopausal status are not entirely clear. The survival difference among premenopausal patients persists even after adjustment for hormone receptor status, lymph node involvement, and the site of recurrent disease. Previous studies have clearly demonstrated that younger age is associated with a poor prognosis after adjustment for known prognostic factors.³ Clearly, breast cancer is a heterogenous disease with differing biological and clinical patterns among younger patients. Early recurrence may represent an indication of adverse biological and clinical features beyond those measured here or even unrecognized at the present time. Classification by disease-free interval may separate patients based on these known and unknown prognostic factors into more and less favorable groups. Likewise, the reason for a lack of prognostic significance of the disease-free interval among postmenopausal women is not certain. Previous studies have demonstrated that older patients have a more favorable clinical course based on time to recurrence and survival.³ There may be less heterogeneity among older women with breast cancer with slower growth rates or spread of disease based on age, hormonal factors, or the biology of the disease. Detailed information on treatments administered subsequent to recurrence was not reviewed, but it is unlikely to differ after consideration of menopausal and hormone receptor status. Whatever the reason, the disease-free interval appears to provide no additional prognostic information relative to subsequent survival among postmenopausal women. As a group, survival subsequent to recurrence in postmenopausal women is as favorable as the survival among premenopausal women who recur late.

Conclusions

The results presented in this study may be of value to clinicians in evaluating patients and estimating prognosis. The application of more effective therapeutic options among premenopausal women with early recurrence should be considered. Alternatively, more conservative management may be reasonable in postmenopausal women and in premenopausal women with late recurrence of disease. These results should also stimulate further research into the relationship between disease-free interval and the biological and clinical parameters that determine these differences in outcome, particularly among premenopausal women.

References

1. Parker SL, Tong T, Bolden S, et al. Cancer statistics, 1997. CA Cancer J Clin. 1997;47:5-27.
2. Miller BA, Ries LA, Hankey BF, et al, eds. SEER Cancer Statistics Review: 1973-1990. Bethesda, Md: National Cancer Institute; NIH Publication No. 93-2789: 1993.
3. Lyman GH, Lyman SL, Balducci L, et al. Age and the risk of breast cancer recurrence. Cancer Control: JMCC. 1996;3:421-427.
4. Lyman GH, Kuderer NM, Lyman SL, et al. Importance of race on breast cancer survival. Ann Surg Oncol. 1997;4:80-87.
5. Wilcoxon E. Individual comparisons by ranking methods. Biomet Bull. 1945;1:80-83.
6. Gart JJ, Thomas DG. The performance of three approximate confidence limits for the odds ratio. Am J Epidemiol. 1982;115:453-470.
7. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457-481.
8. Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep. 1966;50:163-170.
9. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomized trials requiring prolonged observation of each patient, II: analysis and examples. Br J Cancer. 1977;35:1-39.
10. Cox D. Regression models and life tables. J R Stat Soc. 1972;34:187-220.
11. Bartolucci AA, Fraser FD. Comparative step-upward composite tests for selecting prognostic indicators associated with survival. Biomet J. 1977;19:437-438.

From the departments of Internal Medicine (GHL, LB, JH), Surgery (SM, DR, CC), and Epidemiology and Biostatistics (GHL, SLL) at H. Lee Moffitt Cancer Center & Research Institute, Tampa, Fla, and the Albert Ludwigs Universität (NMK, MD), Freiburg, Germany.

Address reprint requests to Dr G. Lyman at the Department of Internal Medicine, H. Lee Moffitt Cancer Center & Research Institute, 12902 Magnolia Dr, Tampa, FL 33612.

Back to Cancer Control Journal Volume 4 Number 4