Age and axillary lymph node ratio in postmenopausal women with T1-T2 node positive breast cancer

Abstract

Purpose: The purpose of this article was to examine the relationship between age and lymph node ratio (LNR, number of positive nodes divided by number of examined nodes), and to determine their effects on breast cancer (BC) and overall mortality.

Methods: Women aged ≥50 years, diagnosed in 1988-1997 with a unilateral histologically confirmed T1-T2 node positive surgically treated primary nonmetastatic BC, were selected from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER). Generalized Additive Models for Location Scale and Shape (GAMLSS) were used to evaluate the age-LNR relationship. Cumulative incidence functions and multivariate competing risks analysis based on model selection by the Bayesian Information Criterion (BIC) were used to examine the effect of age and LNR on mortality. Low LNR was defined as ≤0.20, mid-LNR 0.21-0.65, and high LNR >0.65.

Results: GAMLSS showed a nonlinear LNR-age relationship, increasing from mean LNR 0.26-0.28 at age 50-70 years to 0.30 at 80 years and 0.40 at 90 years. Compared with a 9.8% [95% confidence interval (CI) 8.8%-10.8%] risk of BC death at 5 years in women aged 50-59 years with low LNR, the risk in women ≥80 years with low LNR was 12.6% [95% CI 10.1%-15.0%], mid-LNR 18.1% [13.9%-22.1%], high LNR 29.8% [22.7%-36.1%]. Five-years overall risk of death increased from 40.8% [37.5%-43.9%] by low LNR to 67.4% [61.4%-72.4%] by high LNR. The overall mortality hazard ratio for age ≥80 years with high LNR was 7.49 [6.54-8.59], as compared with women aged 50-59 years with low LNR.

Conclusion: High LNR combined with older age was associated with a threefold increased risk of BC death and a sevenfold increased hazard ratio of overall mortality.

Figure 1.

The distribution of the number of positive nodes, the number of examined nodes, and the LNR, without taking into account age. Unadjusted histograms. Red curves are continuous distribution functions fitted by GAMLSS: lognormal for npos, normal for ntot, and lognormal for LNR. Abbreviations: LNR, lymph node ratio; npos, number of positive nodes; ntot, number of examined nodes.

Figure 2.

The distribution of the number of positive nodes, the number of nodes examined, and the LNR, as a function of age at diagnosis. The three panels represent the distribution of npos, ntot, and LNR according to age. Two representations are combined in each panel, on one hand crude unadjusted observed data, represented by gray and blue dots, and on the other hand semiparametric smoothing by GAMLSS, represented by continuous curves. The gray dots are jittered individual patients' lymph node values, and the large blue dots are the average lymph node values computed separately at each age. The curves are respectively the mean (plain curve in red), the median (plain curve in black), the 25th and 75th percentile (lower and upper dashed curve in black), and the 10th and 90th percentile (lower and upper dotted curve in black). In the ntot panel, the mean and the median curves are identical. Abbreviations: LNR, lymph node ratio; npos, number of positive nodes; ntot, number of examined nodes.

Figure 3.

Cause-specific mortality according to age and LNR. The two panels represent the estimated cause-specific probabilities of dying, either from breast cancer (A) or from other causes (B). The curves are coded by type to indicate the LNR group (plain = high-risk LNR; dashed = intermediary-risk LNR; dotted = low-risk LNR) and by color to indicate the age group (red = age ≥80 years, purple = 70–79 years, green = 60–69 years, and blue = 50–59 years). The same coding applies to the two panels, for example, the red plain curves represent age ≥80 years with high LNR and the purple dashed curves represent age 70–79 years with intermediary LNR. (A): Apart a late deviation of the plain red curve, all curves of the same type are clustered together irrespective of their colors, indicating that patients with the same LNR have comparable risks of dying from breast cancer, regardless of age. (B): Curves of the same color are clustered together irrespective of their types, indicating that patients in the same age group have comparable risks of dying from other causes, regardless of LNR values. The overall mortality can be computed from the two panels by adding the cause-specific mortalities. For example, age 50–59 years with high-risk LNR is represented by the plain blue curves. At 5 years the corresponding breast cancer mortality reads 0.39, the other causes mortality reads 0.04, and the resulting 5-year overall risk of dying is 0.43 (see also Tables 2A and 2B). Abbreviation: LNR, lymph node ratio.