Prognostic significance of mammographic density change after initiation of tamoxifen for ER-positive breast cancer

Abstract

Background: A prior analysis of postmenopausal breast cancer patients linked a decline in mammographic density (MD) following the initiation of tamoxifen treatment with improved survival, but excluded premenopausal women, for whom tamoxifen is the primary anti-endocrine therapy. Therefore, we evaluated change in MD after tamoxifen and breast cancer death among patients age 32 to 87 years.

Methods: This case-control study included 349 estrogen receptor (ER)-positive breast cancer patients who were treated with tamoxifen at Kaiser Permanente Northwest (1990-2008): 97 who died from breast cancer (case patients) and 252 who did not (control patients), matched on age and year at diagnosis and disease stage. Percent MD in the unaffected breast was measured at baseline (mean six months before tamoxifen initiation) and follow-up (mean 12 months after initiation). Associations between change in MD and breast cancer death were estimated using conditional logistic regression.

Results: Patients in the highest tertile of MD decline had a lower risk of breast cancer death when compared with women in the lowest tertile (odds ratio [OR] = 0.44, 95% confidence interval [CI] = 0.22 to 0.88); results were similar after adjustment for baseline MD (OR = 0.49, 95% CI = 0.23 to 1.02). Reductions in death were observed only among patients in the middle and upper tertiles of baseline MD. Associations did not differ by age, tamoxifen use duration, estrogen and/or progestin use, body mass index, or receipt of chemotherapy or radiotherapy.

Conclusion: These data suggest that younger and older ER-positive breast cancer patients who experience large reductions in MD following tamoxifen initiation have an improved prognosis.

Figure 1.

Inclusion and exclusion criteria for study participants. Case patients and a sample of matched control patients were selected from Kaiser Permanente Northwest members diagnosed with estrogen receptor–positive breast cancer and treated with tamoxifen. KPNW = Kaiser Permanente Northwest.

Figure 2.

Selection of baseline and follow-up mammograms. We reviewed mammographic records from January 1, 1988 to December 31, 2010 to select mammographic films of the contralateral (unaffected) breast for analysis. Baseline mammograms were performed prior to the initiation of tamoxifen treatment and follow-up mammograms were obtained at least three months after the start of tamoxifen treatment.

Figure 3.

The association between absolute change in mammographic percent density and breast cancer–specific death, stratified by age at diagnosis. Odds ratios (solid squares) and 95% confidence intervals (error bars) for the association between tertiles of change in percent density and breast cancer death, adjusted for baseline percent density, were similar for patients age 50 years and younger (29 case patients, 73 control patients) and older than age 50 years (68 case patients, 179 control patients) at diagnosis. There was no evidence of multiplicative interaction between age at diagnosis and change in percent density, which was tested by including a cross-product term between the two variables in a logistic regression model (two-sided P _interaction = .96).

Figure 4.

The magnitude of absolute change in percent density among case patients (A) and control patients (B) who had gaps between tamoxifen prescriptions of 31 to 60, 61 to 90, or more than 90 days in between baseline and follow-up mammograms is shown in comparison with patients who did not have any gaps of at least 30 days during the same time period. Thick horizontal bars represent the median value in each group. Upper whiskers extend from the 75^th percentile to the highest value within 1.5 times the interquartile range, and lower whiskers extend from the 25^th percentile to the lowest value within 1.5 times the interquartile range. A minority of patients were nonadherent prior to the follow-up mammogram: 15.5% of case patients and 13.1% of control patients had a gap of 31 to 60 days, 4.1% of case patients and 3.6% of control patients had a gap of 61 to 90 days, and 5.2% of case patients and 4.0% of control patients had a gap of more than 90 days. Levels of absolute change in percent density did not differ statistically significantly by nonadherence among case patients (A, Kruskal Wallis test, two-sided P = .06) or control patients (B, Kruskal Wallis test, two-sided P = .18).