Purpose:

The choice of therapy for patients with breast cancer is often based on clinicopathologic parameters, hormone receptor status, and HER2 amplification. To improve individual prognostication and tailored treatment decisions, we combined clinicopathologic prognostic data with genome instabilty profiles established by quantitative measurements of the DNA content.

Experimental Design:

We retrospectively assessed clinical data of 4,003 patients with breast cancer with a minimum postoperative follow-up period of 10 years. For the entire cohort, we established genome instability profiles. We applied statistical methods, including correlation matrices, Kaplan–Meier curves, and multivariable Cox proportional hazard models, to ascertain the potential of standard clinicopathologic data and genome instability profiles as independent predictors of disease-specific survival in distinct subgroups, defined clinically or with respect to treatment.

Results:

In Cox regression analyses, two parameters of the genome instability profiles, the S-phase fraction and the stemline scatter index, emerged as independent predictors in premenopausal women, outperforming all clinicopathologic parameters. In postmenopausal women, age and hormone receptor status were the predominant prognostic factors. However, by including S-phase fraction and 2.5c exceeding rate, we could improve disease outcome prediction in pT1 tumors irrespective of the lymph node status. In pT3-pT4 tumors, a higher S-phase fraction led to poorer prognosis. In patients who received adjuvant endocrine therapy, chemotherapy or radiotherapy, or a combination, the ploidy profiles improved prognostication.

Conclusions:

Genome instability profiles predict disease outcome in patients with breast cancer independent of clinicopathologic parameters. This applies especially to premenopausal patients. In patients receiving adjuvant therapy, the profiles improve identification of high-risk patients.

Translational Relevance

In this study, we collected 4,003 primary invasive breast carcinomas over a period of 13 years with a minimum postoperative follow-up time of 10 years. By using image cytometry, we determined seven different parameters to generate genome instability profiles of the tumors. We show that genome instability profiles have additional prognostic value independent of standard clinical prognostic factors, especially in the group of premenopausal patients. Genome instability profiles of breast cancers can, when combined with standard clinicopathologic features, improve prognostication, and thus contribute to individualized medicine.

Breast cancer is the most common cancer among women and, with more than 620,000 deaths worldwide in 2018, also the major cause of cancer-related deaths (1). In addition to surgery and radiotherapy, systemic adjuvant treatment regimens were developed, including chemotherapy and endocrine therapy. Intensive research in the last decades has resulted in the identification of a few biomarkers for therapy stratification, some of which are now included in standard clinical guidelines (2). Estrogen receptor (ER) and progesterone receptor (PR) expression has been routinely measured for more than 30 years and most patients with hormone receptor–positive cancers benefit from antihormonal drugs (3–5). Breast cancers with overexpression of HER2, accounting for around 20% of the cases, can be treated with targeted antibody therapy (6). In addition, Ki-67 expression as a marker of proliferative activity is used as an indication for chemotherapy; however, it is not always part of the diagnostic procedures (7). Therefore, in routine diagnostics, the decision for adjuvant chemotherapy is often based on the estimated prognosis of the patient and additional biomarkers to improve the prognostication or predict therapy response are urgently needed.

Global gene expression analyses were used to calculate the risk for, for example, disease recurrence or therapy failure. Oncotype DX (8) and MammaPrint (9) are two of the most commonly used tests. However, not all patients with breast cancer can benefit: MammaPrint is only valid for early-stage breast cancer (lymph node stage N0-N1), and Oncotype DX is solely applicable for patients with ER-positive and HER2-negative tumors (8–10). In 2009, Swanton and colleagues developed a signature of genes that are linked to chromosomal instability and showed that a high expression of these genes was significantly associated with a poorer disease-free and disease-specific survival (DSS) rate of patients with breast cancer (11). Habermann and colleagues described a gene expression signature of genomic instability in patients with breast cancer (12). The authors showed that both MammaPrint and Oncotype DX expression signatures correlated with the degree of genomic instability (13), determined by quantitative measurements of the nuclear DNA content of the tumor cells using image cytometry. Dozens of studies have been published to finally clarify the prognostic value of the DNA content measurements (14). However, due to (i) methodologic differences between the studies, such as different classification criteria, (ii) the inclusion of different cytometric parameters, and (iii) heterogeneous patient cohorts, contradictory results emerged (15–19).

We now present a comprehensive comparison of the predictive power of routine clinicopathologic prognostic parameters with genome instability profiles established using image cytometry by applying objective, quantitative parameters in 4,003 patients with breast cancer with at least 10 years clinical follow-up.

Study cohort and clinicopathologic data

Our cohort consists of 4,003 female patients with breast cancer diagnosed in the greater Stockholm area (Karolinska University Hospital, Vällingby Clinical Centre as well as other outpatient clinics in Stockholm, Sweden) between 1988 and 2001, for whom clinicopathologic and image cytometry data were available. The mean age of the patients was 59 years (range: 23–95). We retrospectively collected survival data over a postoperative period of at least 10 years (median follow-up 13.2 years). Data acquisition was closed in December 2012. The study was conducted in compliance with the Declaration of Helsinki. The local Swedish ethics committee approved the use of anonymized data for this retrospective study without contacting the patients (Dnr 2013/707-31/3). All clinicopathologic parameters that were included in the analyses are listed in Table 1. Pathologic tumor and lymph node stages were classified according to Union for International Cancer Control guidelines (20). The threshold for ER and PR positivity was set at ≥0.05 fmol/ng DNA, which was the standard procedure at Karolinska Hospital (Stockholm, Sweden) during that time period. Proliferative activity of the tumor cells was assessed by measurement of either Ki-67 or Cyclin A levels by immunocytochemistry. Tumors were classified as highly proliferating if more than 5% and 10% cells were positive for Cyclin A and Ki-67, respectively. As for most patients only one of these parameters had been assessed during routine diagnostics, we created a combined proliferation variable, which indicated high proliferation if at least one of the two parameters had reached the respective threshold.

Table 1.

Kaplan–Meier and univariable Cox analyses for clinicopathologic and genome instability parameters.

VariableCategoriesN [Patients (%)]Mean DSS95% CIHR DSS95% CIP log-rank KPM, DSS
Standard clinicopathologic data 
Age at diagnosis in years ≤40 255 (6.4) 16.4 15.5–17.3 ref  <0.001 
 41–70 2,691 (67.2) 20.0 19.7–20.2 0.58 0.45–0.74  
 >70 1,057 (26.4) 16.4 15.9–16.9 1.03 0.79–1.34  
Menopausal status Premenopausal 1,057 (26.4) 18.7 18.2–19.1 ref  0.071 
 0–5 years postmenopausal 450 (11.2) 18.5 17.9–19.1 0.81 0.62–1.05  
 >5 years postmenopausal 2,320 (58.0) 17.8 17.5–18.1 1.07 0.91–1.25  
 Unknown 176 (4.4) – – – –  
Recurrences No 3,442 (86.0) 20.1 19.9–20.4 ref  <0.001 
 Yes 561 (14.0) 14.0 13.3–14.7 3.46 2.98–4.02  
Metastases No 3,159 (78.9) 22.3 22.1–22.4 ref  <0.001 
 Yes 844 (21.1) 8.9 8.4–9.4 32.99 27.2–40.0  
Tumor stage pT1 2,327 (58.1) 19.8 19.5–20.0 ref  <0.001 
 pT2 1,281 (32.0) 17.4 16.9–17.9 2.41 2.01–2.80  
 pT3-pT4 152 (3.8) 13.3 11.8–14.7 4.26 3.26–5.57  
 Unknown 243 (6.1) – – – –  
Lymph node stage pN0 2,203 (55.0) 19.6 19.4–19.9 ref  <0.001 
 pN1 968 (24.2) 18.6 18.1–19.1 2.10 1.76–2.50  
 pN2-pN3 475 (11.9) 12.6 11.8–13.4 5.65 4.74–6.74  
 Unknown 357 (8.9) — — — —  
Lymph node stage (dichotomized) pN0 2,203 (55.0) 19.6 19.4–19.9 ref  <0.001 
 pN1-pN3 1,443 (36.0) 16.9 16.4–17.3 3.07 2.64–3.57  
 Unknown 357 (8.9) — — — —  
Progesterone receptor status Positive 2,584 (64.6) 20.1 19.8–20.4 ref  <0.001 
 Negative 1,052 (26.3) 16.1 15.6–16.6 2.04 1.76–2.37  
 Unknown 367 (9.2) — — — —  
Estrogen receptor status Positive 2,823 (70.5) 19.9 19.6–20.2 ref  <0.001 
 Negative 825 (20.6) 16.0 15.4–16.6 1.94 1.66–2.27  
 Unknown 355 (8.9) — — — —  
Genome instability data 
S-phase fraction 0–5% 2,525 (63.1) 19.2 19.0–19.4 ref  <0.001 
 >5% 1,364 (34.1) 17.0 16.6–17.5 2.12 1.84–2.45  
 Unknown 114 (2.8) — — — —  
G2–M exceeding rate 0–1% 3,312 (82.7) 18.1 17.9–18.4 ref  <0.001 
 >1% 660 (16.5) 16.2 15.6–16.8 1.67 1.41–1.98  
 Unknown 31 (0.8) — — — —  
2.5c exceeding rate <5% 803 (20.1) 19.3 19.0–19.7 ref  <0.001 
 5–50% 1,269 (31.7) 18.1 17.8–18.5 1.85 1.44–2.39  
 >50% 1,900 (47.5) 16.9 16.6–17.3 2.66 2.10–3.37  
 Unknown 31 (0.8) — — — —  
5c exceeding rate 0% 1,451 (36.2) 19.0 18.7–19.3 ref  <0.001 
 0.1–10% 1,547 (38.6) 17.8 17.5–18.1 1.65 1.37–1.98  
 >10% 977 (24.4) 16.0 15.5–16.5 2.69 2.23–3.24  
 Unknown 28 (0.7) — — — —  
Stemline ≤2.2c 1,612 (40.3) 18.8 18.5–19.1 ref  <0.001 
 2.21–3.8c 1,472 (36.8) 18.3 17.9–18.8 1.94 1.64–2.30  
 3.81–4.2c 624 (15.6) 17.7 17.2–18.2 1.53 1.23–1.91  
 >4.2c 295 (7.4) 16.6 15.7–17.5 2.24 1.73–2.89  
Coefficient of variation (tumor cells) 0–5% 2,786 (69.6) 18.0 17.7–18.2 ref  0.003 
 >5% 1,189 (29.7) 17.3 16.9–17.7 1.25 1.08–1.46  
 Unknown 28 (0.7) — — — —  
Stemline scatter index Low (≤8.8) 2,159 (53.9) 18.7 18.5–18.9 ref  <0.001 
 High (>8.8) 1,704 (42.6) 16.7 16.3–17.1 1.94 1.68–2.25  
 Unknown 140 (3.5) — — — —  
VariableCategoriesN [Patients (%)]Mean DSS95% CIHR DSS95% CIP log-rank KPM, DSS
Standard clinicopathologic data 
Age at diagnosis in years ≤40 255 (6.4) 16.4 15.5–17.3 ref  <0.001 
 41–70 2,691 (67.2) 20.0 19.7–20.2 0.58 0.45–0.74  
 >70 1,057 (26.4) 16.4 15.9–16.9 1.03 0.79–1.34  
Menopausal status Premenopausal 1,057 (26.4) 18.7 18.2–19.1 ref  0.071 
 0–5 years postmenopausal 450 (11.2) 18.5 17.9–19.1 0.81 0.62–1.05  
 >5 years postmenopausal 2,320 (58.0) 17.8 17.5–18.1 1.07 0.91–1.25  
 Unknown 176 (4.4) – – – –  
Recurrences No 3,442 (86.0) 20.1 19.9–20.4 ref  <0.001 
 Yes 561 (14.0) 14.0 13.3–14.7 3.46 2.98–4.02  
Metastases No 3,159 (78.9) 22.3 22.1–22.4 ref  <0.001 
 Yes 844 (21.1) 8.9 8.4–9.4 32.99 27.2–40.0  
Tumor stage pT1 2,327 (58.1) 19.8 19.5–20.0 ref  <0.001 
 pT2 1,281 (32.0) 17.4 16.9–17.9 2.41 2.01–2.80  
 pT3-pT4 152 (3.8) 13.3 11.8–14.7 4.26 3.26–5.57  
 Unknown 243 (6.1) – – – –  
Lymph node stage pN0 2,203 (55.0) 19.6 19.4–19.9 ref  <0.001 
 pN1 968 (24.2) 18.6 18.1–19.1 2.10 1.76–2.50  
 pN2-pN3 475 (11.9) 12.6 11.8–13.4 5.65 4.74–6.74  
 Unknown 357 (8.9) — — — —  
Lymph node stage (dichotomized) pN0 2,203 (55.0) 19.6 19.4–19.9 ref  <0.001 
 pN1-pN3 1,443 (36.0) 16.9 16.4–17.3 3.07 2.64–3.57  
 Unknown 357 (8.9) — — — —  
Progesterone receptor status Positive 2,584 (64.6) 20.1 19.8–20.4 ref  <0.001 
 Negative 1,052 (26.3) 16.1 15.6–16.6 2.04 1.76–2.37  
 Unknown 367 (9.2) — — — —  
Estrogen receptor status Positive 2,823 (70.5) 19.9 19.6–20.2 ref  <0.001 
 Negative 825 (20.6) 16.0 15.4–16.6 1.94 1.66–2.27  
 Unknown 355 (8.9) — — — —  
Genome instability data 
S-phase fraction 0–5% 2,525 (63.1) 19.2 19.0–19.4 ref  <0.001 
 >5% 1,364 (34.1) 17.0 16.6–17.5 2.12 1.84–2.45  
 Unknown 114 (2.8) — — — —  
G2–M exceeding rate 0–1% 3,312 (82.7) 18.1 17.9–18.4 ref  <0.001 
 >1% 660 (16.5) 16.2 15.6–16.8 1.67 1.41–1.98  
 Unknown 31 (0.8) — — — —  
2.5c exceeding rate <5% 803 (20.1) 19.3 19.0–19.7 ref  <0.001 
 5–50% 1,269 (31.7) 18.1 17.8–18.5 1.85 1.44–2.39  
 >50% 1,900 (47.5) 16.9 16.6–17.3 2.66 2.10–3.37  
 Unknown 31 (0.8) — — — —  
5c exceeding rate 0% 1,451 (36.2) 19.0 18.7–19.3 ref  <0.001 
 0.1–10% 1,547 (38.6) 17.8 17.5–18.1 1.65 1.37–1.98  
 >10% 977 (24.4) 16.0 15.5–16.5 2.69 2.23–3.24  
 Unknown 28 (0.7) — — — —  
Stemline ≤2.2c 1,612 (40.3) 18.8 18.5–19.1 ref  <0.001 
 2.21–3.8c 1,472 (36.8) 18.3 17.9–18.8 1.94 1.64–2.30  
 3.81–4.2c 624 (15.6) 17.7 17.2–18.2 1.53 1.23–1.91  
 >4.2c 295 (7.4) 16.6 15.7–17.5 2.24 1.73–2.89  
Coefficient of variation (tumor cells) 0–5% 2,786 (69.6) 18.0 17.7–18.2 ref  0.003 
 >5% 1,189 (29.7) 17.3 16.9–17.7 1.25 1.08–1.46  
 Unknown 28 (0.7) — — — —  
Stemline scatter index Low (≤8.8) 2,159 (53.9) 18.7 18.5–18.9 ref  <0.001 
 High (>8.8) 1,704 (42.6) 16.7 16.3–17.1 1.94 1.68–2.25  
 Unknown 140 (3.5) — — — —  

Abbreviations: CI, confidence interval; DSS, disease-specific survival; HR, hazard ratio; KPM, Kaplan–Meier; ref, reference category.

All patients included in the study were treated rule-based with mastectomy or partial mastectomy, and, if indicated combined with postoperative local radiotherapy. Patients with ER–positive carcinomas received adjuvant endocrine therapy (tamoxifen and/or zoladex). Patients with advanced disease were treated with adjuvant chemotherapy (cyclophosphamide, methotrexate, fluorouracil or anthracycline-based).

Image cytometry

Fine needle aspirates (FNA) were prospectively collected from all patients and directly processed for diagnostic purposes. FNAs were stained according to Feulgen, following standard procedures to quantify the DNA content of the nuclei (13, 21, 22). Normal lymphocytes served as controls. In every sample, between 100 and 200 tumor cells were analyzed. Examples of different genome instability profiles are shown in Supplementary Fig. S1. The analysis included the measurement of six different parameters (Table 2). In addition, as a seventh parameter, we assessed the “stemline scatter index” (SSI) according to Kronenwett and colleagues as the sum of the percentage of cells in the S-phase fraction, the G2–M exceeding rate and the coefficient of variation of the tumor cells and considered values ≤8.8 as SSI low and >8.8 as SSI high, as described previously (21).

Table 2.

Genome instability parameters assessed by image cytometry.

ParameterDescription
Stemline The stemline value describes the DNA content of the cells in the G0–G1-phase, which is generally the highest peak of the histogram. 
 It is denoted in “c” units, whereby, e.g., “2c” indicates a diploid and “4c” a tetraploid stemline. 
2.5c exceeding rate The 2.5c exceeding rate describes the amount of cells with a DNA content >2.5c. 
 It is given in % of all analyzed cells. 
5c exceeding rate The 5c exceeding rate describes the amount of cells with a DNA content 5c. 
 It is given in % of all analyzed cells. 
G2–M exceeding rate The G2–M exceeding rate describes the proportion of cells that show a higher DNA amount than the cells in the G2–M phase of the major cell population. 
 It is not dependent on the stemline value and therefore describes how many cells are seen beyond the major cell population. 
 It is given in % of all analyzed cells. 
S-phase fraction The S-phase fraction describes the proliferating cells that show a DNA content between those of the G0–G1 and G2–M peaks. 
 It is given in % of all analyzed cells. 
Coefficient of variation The coefficient of variation of the tumor cells is calculated by dividing the SD of the measured “c” values of the G0–G1-peak by their mean. 
 It is an indicator for the spread of the G0–G1-peak and therefore provides information about the heterogeneity of the stemline. 
ParameterDescription
Stemline The stemline value describes the DNA content of the cells in the G0–G1-phase, which is generally the highest peak of the histogram. 
 It is denoted in “c” units, whereby, e.g., “2c” indicates a diploid and “4c” a tetraploid stemline. 
2.5c exceeding rate The 2.5c exceeding rate describes the amount of cells with a DNA content >2.5c. 
 It is given in % of all analyzed cells. 
5c exceeding rate The 5c exceeding rate describes the amount of cells with a DNA content 5c. 
 It is given in % of all analyzed cells. 
G2–M exceeding rate The G2–M exceeding rate describes the proportion of cells that show a higher DNA amount than the cells in the G2–M phase of the major cell population. 
 It is not dependent on the stemline value and therefore describes how many cells are seen beyond the major cell population. 
 It is given in % of all analyzed cells. 
S-phase fraction The S-phase fraction describes the proliferating cells that show a DNA content between those of the G0–G1 and G2–M peaks. 
 It is given in % of all analyzed cells. 
Coefficient of variation The coefficient of variation of the tumor cells is calculated by dividing the SD of the measured “c” values of the G0–G1-peak by their mean. 
 It is an indicator for the spread of the G0–G1-peak and therefore provides information about the heterogeneity of the stemline. 

Statistical analyses

If not indicated otherwise, statistical analyses were conducted in IBM SPSS Statistics V22.0. The relationship between all clinicopathologic and image cytometric data was assessed by Spearman rank correlation coefficient and visualized by a matrix heatmap.

For survival analyses, DSS was used as endpoint. DSS is defined as the time until death due to breast cancer has occurred. We generated survival curves using the Kaplan–Meier method and tested equivalency of the survival distributions by the log-rank test. Univariable Cox regressions were conducted, estimating the HRs with 95% confidence intervals (CI). For all survival analyses, we categorized continuous variables according to the survival curves taking into account the group size to ensure robust statistical analyses.

Multivariable Cox regression models were used to evaluate the independence of standard clinicopathologic or image cytometric data as predictors for the DSS. We included all variables that had been analyzed in univariable analyses as potential predictors. The variables “recurrences” and “metastases” were excluded because they cannot be used for a priori prognostication. Only cases without any missing values were included in the models. The number of available cases as well as the number of events, is shown for each model separately next to the forest plot in the Results section.

All multivariable Cox regressions were performed using the backward Wald method and variables were retained with P < 0.05 (23, 24). Multivariable analyses were conducted for 16 subgroups of patients separately. On the basis of clinical and pathologic data of patients and tumors, we created nine subgroups (Fig. 1A): patients with tumor stages pT1 and pT2 were differentiated into premenopausal and postmenopausal and further subdivided according to lymph node stage (pN0 and pN1–pN3). This resulted in eight subgroups. Patients with tumor stage pT3 and pT4 were analyzed as one group (N = 113) due to the limited case number. We formed another seven subgroups consisting of patients who had received (1) only endocrine therapy, (2) only radiotherapy, (3) only chemotherapy, or (4)–(7) a combination of these therapies (Fig. 1B). For every subgroup, the initial model as well as the steps to the final model are provided in Supplementary Table S1.

Figure 1.

Subgroups for Cox analyses. Gray boxes indicate the analyzed subgroups. A, We formed nine pathologic subgroups. For better clarity, the groups were given the listed names. Low-stage tumors were subgrouped by menopausal stage, tumor stage, and lymph node stage. High-stage tumors (pT3 and pT4) could not further be subgrouped due to the small case number. B, For the analyses regarding therapy schemes, we formed seven subgroups of patients, who received different adjuvant treatment regimens. CT, chemotherapy; endo, endocrine therapy; RT, radiotherapy.

Figure 1.

Subgroups for Cox analyses. Gray boxes indicate the analyzed subgroups. A, We formed nine pathologic subgroups. For better clarity, the groups were given the listed names. Low-stage tumors were subgrouped by menopausal stage, tumor stage, and lymph node stage. High-stage tumors (pT3 and pT4) could not further be subgrouped due to the small case number. B, For the analyses regarding therapy schemes, we formed seven subgroups of patients, who received different adjuvant treatment regimens. CT, chemotherapy; endo, endocrine therapy; RT, radiotherapy.

Close modal

Multivariable analyses including the proliferation data of Ki-67 and Cyclin A (21% of the cases had data for at least one of the markers), were performed within the premenopausal and postmenopausal patient groups, not subgrouped any further.

After Cox regression analyses, for each model, we generated a model excluding the instability parameters that had remained as predictors. We then compared the models with and without genome instability using a likelihood ratio test to ascertain whether the genome instability indeed provides additional predictive power. The test was conducted using the function “anova” in R.

For the subsequent development of the nomograms, two further Cox analyses were performed, including all premenopausal and postmenopausal patients, respectively. The models were internally validated by 1,000 bootstrap resamples and the optimism-corrected concordance index (C-Index) was calculated. The nomograms were developed using the nomogram package in R. We integrated all parameters that remained independent predictors in the Cox models of the subgroups. For the nomograms, age and image cytometric data were included as continuous variables for a better visualization.

The proportional hazards assumption was checked for each variable in all Cox models using the Schoenfeld statistical test in R (package “survival”; ref. 25). The assumption could be supported except for five variables in four different subgroups. Kaplan–Meier curves of these variables and the plots of the Schoenfeld residuals are shown in Supplementary Fig. S2.

The complete available data in this study with more than 4,000 cases had a power above 98% to detect a HR of 1.3 in a Cox regression analysis (two-sided significance level 0.05, group ratio 1:1, assumed survival rate 0.8 vs. 0.9). However, the sample size of the actual conducted subset analyses with a mean of 354 patients (ranging from 96 to 987) achieved the average power of 75% for a presumed HR of 2.11.

All reported P values are two-sided.

We retrospectively evaluated a cohort of 4,003 patients to investigate to which extent genome instability profiles could improve prediction of disease outcome in women with breast cancer. For the entire cohort, the nuclear DNA content had been quantitatively measured to generate genome instability profiles. The clinical follow-up was at least 10 years.

The results of Kaplan–Meier and univariable Cox analyses regarding clinicopathologic parameters were consistent with the literature (26–28; Table 1; Supplementary Figs. S3 and S4).

To establish genome instability profiles, we measured six parameters of which all were associated with a shorter DSS when increased (Table 1; Supplementary Figs. S3 and S4). These include the S-phase fraction (>5%: HR 2.12; 95% CI, 1.84–2.45), the G2–M exceeding rate (>1%: HR 1.67; 95% CI, 1.41–1.98), the 2.5c exceeding rate (5–50%: HR 1.85; 95% CI, 1.44–2.39 and >50%: HR 2.66; 95% CI, 2.10–3.37), the 5c exceeding rate (0.1%–10%: HR 1.65; 95% CI, 1.37–1.98 and >10%: HR 2.69; 95% CI, 2.23–3.24), the stemline value (2.21–3.8c: HR 1.94; 95% CI, 1.64–2.30; 3.81–4.2c: HR 1.53; 95% CI, 1.23–1.91 and >4.2c: HR 2.24; 95% CI, 1.73–2.89), all with P < 0.001, and the coefficient of variation of the tumor cells (>5%: HR 1.25; 95% CI, 1.08–1.46; P = 0.003).

Likewise, the SSI (see Materials and Methods) was associated with DSS (>8.8: HR 1.94; 95% CI, 1.68–2.25; P < 0.001).

High genome instability profiles are correlated with hormone receptor negativity

We assessed possible associations between the parameters that determine the genome instability profiles and clinicopathologic data (Supplementary Table S2). Here, ER and PR positivity showed the strongest correlation coefficient with the following parameters: a low 5c exceeding rate (−0.324 and −0.299), a SSI ≤8.8 (−0.303 and −0.281), a low S-phase fraction (−0.294 and −0.280), a low G2–M exceeding rate (−0.271 and −0.223), and a low 2.5 exceeding rate (−0.223 and −0.224), that is, receptor-positive tumors show a lower degree of genomic instability.

S-phase fraction and SSI are independent prognostic factors in premenopausal patients

In premenopausal patients, genome instability profiles were the dominant independent predictors of DSS in multivariable Cox regression models (Fig. 2). In the T1N0 and T2N+ subgroups, the S-phase fraction was the only remaining predictor of DSS (HR 2.02; 95% CI, 1.08–3.79 and HR 1.74; 95% CI, 1.04–2.91, respectively). The SSI was predictive in the T1N+ and T2N0 subgroups (HR 2.05; 95% CI, 1.07–3.92 and HR 4.05; 95% CI, 1.21–13.59, respectively).

Figure 2.

Multivariable Cox regression models of clinicopathologic subgroups. For Cox regression models, we formed nine subgroups of patients, based on the pathologic features that are commonly used for prognostication (menopausal status, tumor, and lymph node stage). All standard clinicopathologic and image cytometric data that are available at time of diagnosis or surgery were included in the proportional hazard model. Independent prognostic factors were selected with P < 0.05. N indicates the number of available cases (+ number of events) for every analysis. Clinicopathologic features are shown in gray, genome instability parameters in black. The dashed line indicates a HR of 1, the error bars show the 95% CIs. CI, Confidence interval; ER, estrogen receptor; HR, hazard ratio; PR, progesterone receptor; ref, reference category.

Figure 2.

Multivariable Cox regression models of clinicopathologic subgroups. For Cox regression models, we formed nine subgroups of patients, based on the pathologic features that are commonly used for prognostication (menopausal status, tumor, and lymph node stage). All standard clinicopathologic and image cytometric data that are available at time of diagnosis or surgery were included in the proportional hazard model. Independent prognostic factors were selected with P < 0.05. N indicates the number of available cases (+ number of events) for every analysis. Clinicopathologic features are shown in gray, genome instability parameters in black. The dashed line indicates a HR of 1, the error bars show the 95% CIs. CI, Confidence interval; ER, estrogen receptor; HR, hazard ratio; PR, progesterone receptor; ref, reference category.

Close modal

Genomic instability profiles and clinicopathologic data have prognostic value in postmenopausal patients

Compared with the premenopausal subgroups, genome instability profiles played a less dominant role in postmenopausal patients (Fig. 2). However, a higher S-phase fraction was, in addition to age (>70) and PR negativity, an independent predictor in the T1N+ subgroup (HR 2.43; 95% CI, 1.54–3.83); the 2.5c exceeding rate remained an independent predictor in T1N0 patients (5–50%: HR 4.42; CI, 2.05–9.51 and >50%: HR 3.37; 95% CI, 1.56–7.28).

S-phase fraction is an independent predictor in high tumor stages

Due to the limited case number (N = 113), we fitted only one regression model comprising both pT3 and pT4 tumors (Fig. 2). Lymph node positivity showed predictive relevance in the regression model (HR 3.34; 95% CI, 1.48–7.55), and so did age (HR 3.52; 95% CI, 1.31–9.46 for age ≤40 and HR 2.60; 95% CI, 1.27–5.29 for age >70) and ER negativity (HR 2.68; 95% CI, 1.39–5.20). Increased S-phase fraction remained an independent predictor (HR 2.09; 95% CI, 1.13–3.88) for DSS.

Clinicopathologic data and genome instability profiles have predictive power in patients treated with adjuvant therapy

We tested whether genome instability parameters could improve prognostication vis-à-vis specific treatments. We therefore computed regression models of seven therapeutic subgroups separately (Fig. 1B). In five of the subgroups (i.e., endocrine, radiotherapy, endocrine + radiotherapy, radiotherapy + chemotherapy, endocrine + radiotherapy + chemotherapy), prognostication could be improved by adding genome instability parameters in the model (Fig. 3; Supplementary Table S3). For patients treated with chemotherapy alone, the genome instability parameters did not show independent prognostic impact. For patients treated with endocrine and chemotherapy, no model could be fitted.

Figure 3.

Multivariable Cox regression models of therapeutic subgroups. For further Cox regressions, we formed seven subgroups of patients who had received similar adjuvant treatment. The analysis of patients, who received a combination of chemotherapy and endocrine therapy, did not reveal any significant predictor and was therefore not displayed. N indicates the number of available cases (+ number of events) for every analysis. Clinicopathologic features are shown in gray, genome instability parameters in black, the error bars show the 95% CIs. The dashed line indicates a HR of 1. CI, confidence interval; ER, estrogen receptor; HR, hazard ratio; PR, progesterone receptor; ref, reference category.

Figure 3.

Multivariable Cox regression models of therapeutic subgroups. For further Cox regressions, we formed seven subgroups of patients who had received similar adjuvant treatment. The analysis of patients, who received a combination of chemotherapy and endocrine therapy, did not reveal any significant predictor and was therefore not displayed. N indicates the number of available cases (+ number of events) for every analysis. Clinicopathologic features are shown in gray, genome instability parameters in black, the error bars show the 95% CIs. The dashed line indicates a HR of 1. CI, confidence interval; ER, estrogen receptor; HR, hazard ratio; PR, progesterone receptor; ref, reference category.

Close modal

Genome instability profiles significantly increase the predictive power in pathologic and therapeutic subgroups

For all Cox analyses in which genome instability parameters remained as predictors in the final model, that is, in seven pathologic and five therapeutic subgroups, we wanted to ascertain if they provide a better prediction compared with the same model without genome instability parameters. The likelihood ratio test showed that for all analyses, the models including genome instability parameters significantly improved the prediction compared with their corresponding models with only standard prognostic parameters (Supplementary Table S3).

The SSI remains an independent predictor in premenopausal patients if analyzed in multivariable analyses with standard proliferation markers

SSI and S-phase fraction were the dominant prognostic factors in multivariable models, both being associated with cell proliferation. Therefore, we wanted to determine if they remained independent predictors within the patients for whom independent proliferation data (Ki-67/Cyclin A) were also available. In premenopausal patients (N = 225), Cox regression determined two variables as independent prognostic factors: pathologic tumor stage (HR 0.73; 95% CI, 0.34–1.57 for pT2 and HR 15.11; 95% CI, 5.27–43.32 for pT3-pT4) and the SSI >8.8 (HR 2.96; 95% CI, 1.42–6.15). The Ki-67/Cyclin A proliferation variable could not improve prognostication in premenopausal patients. In contrast, in postmenopausal patients (N = 617), high Ki-67/Cyclin A levels remained an independent prognostic factor (HR 1.92; 95% CI, 1.30–2.83), while none of the genome instability profile parameters remained independent predictors (Supplementary Fig. S5A). Excluding the proliferation variable from this analysis, we determined that the S-phase fraction remained an independent predictor (>5%: HR 1.73; 95% CI, 1.14–2.62), now substituting the Ki-67/Cyclin A variable (Supplementary Fig. S5B).

Nomograms for the survival probability based on clinicopathologic and genomic instability parameters

On the basis of the independent predictors found in the Cox regression analyses, we developed two nomograms separately for premenopausal and postmenopausal patients (Fig. 4). Nomograms estimate the survival probability for a patient after 5 and 10 years based on a total score which is calculated by addition of zero to 100 points for every individual predictor. In the nomogram for premenopausal patients (Fig. 4A), the S-phase fraction showed the highest impact of all parameters because a high value of the S-phase fraction can add up to 100 points to the final score. In contrast, in postmenopausal patients, the instability parameters (S-phase fraction and 2.5c exceeding rate) only play a minor role for prediction (Fig. 4B). The internal validation of the underlying regression models showed optimism-corrected C-index values of 0.69 and 0.74 for premenopausal and postmenopausal patients, respectively.

Figure 4.

Nomograms for premenopausal (A) and postmenopausal (B) patients. Parameters that were independent predictors in the Cox regression models were used to develop the nomograms. For every parameter, a score on the upper points scale is given. On the basis of the sum of all separate parameters, the 5-year and 10-year survival probability can be estimated by drawing a vertical line from the “Total points” scale. ER, estrogen receptor; PR, progesterone receptor.

Figure 4.

Nomograms for premenopausal (A) and postmenopausal (B) patients. Parameters that were independent predictors in the Cox regression models were used to develop the nomograms. For every parameter, a score on the upper points scale is given. On the basis of the sum of all separate parameters, the 5-year and 10-year survival probability can be estimated by drawing a vertical line from the “Total points” scale. ER, estrogen receptor; PR, progesterone receptor.

Close modal

Our analyses were conducted on a cohort of 4,003 patients with breast cancer that had been analyzed by image cytometry. We first analyzed routine diagnostic parameters by Kaplan–Meier survival analyses using DSS. Our results confirmed the current state of the literature and therefore the integrity of our patient cohort (26–28).

Image cytometric measurements of the nuclear DNA content have been discussed as additional prognostic parameters in several cancer entities, particularly in breast cancer, for more than 30 years (14, 29). In addition, image cytometry can provide information about the proliferation rate or intratumor heterogeneity (30, 31). The degree of aneuploidy and intratumor heterogeneity influences disease outcome (14, 32).

We now analyzed genome instability profiles using image cytometry in a coherent cohort of 4,003 patients with breast cancer to assess its prognostic potential. In Kaplan–Meier analyses and univariable Cox regressions, all cytometric parameters showed prognostic significance, and the results were consistent with the underlying biology.

Except for hormone receptor status and age at diagnosis, none of the clinicopathologic parameters showed strong associations with parameters of the genome instability profiles. Therefore, we surmised that the genome instability profiles could have independent prognostic value when analyzed together with clinicopathologic parameters, which we validated in multivariable analyses.

The S-phase fraction and the SSI emerged as the most important image cytometric parameters for prognosis and, particularly in premenopausal patients, even more meaningful than standard clinicopathologic parameters. The S-phase fraction was already subject of dozens of studies, analyzing its prognostic value in breast cancer. Assessed by means of image cytometry in breast cancer, it was shown to correlate with disease-specific survival as well as with larger tumor sizes and younger age at diagnosis (19, 33). However, to our knowledge so far, no study analyzing the S-phase fraction by image cytometry on a comparably large and coherent cohort has been published. In contrast to the S-phase fraction, the SSI, first proposed by Kronenwett and colleagues (21), is less commonly studied. Nevertheless, in our analyses, it showed clear potential as an independent prognostic factor in premenopausal patients.

Because the assessment of the SSI includes the S-phase fraction, both of these parameters are associated with the proliferative activity of the tumor cells. The current approach for diagnosis and therapy decision in breast cancer includes IHC staining of Ki-67 as a proliferation marker. Studies already showed that both the image cytometric and flow cytometric S-phase fraction are significantly associated with the Ki-67 expression of breast tumors (33–35). In our analyses of the subgroup of patients for whom data of the standard proliferation markers Ki-67 or Cyclin A were available, the SSI remained an independent predictor for the prognosis of premenopausal patients, outperforming the standard proliferation variable.

Regarding therapeutic approaches, we could show that genome instability profiles help to detect high-risk patients in five of seven therapeutic subgroups and therefore might be a tool to tailor adjuvant therapy. Patients who were treated with endocrine therapy only had a worse prognosis if they showed high 5c and/or G2–M exceeding rates. In these cases, adjuvant radiotherapy or chemotherapy might have been suitable. This applies, in particular, to young (≤40 years) and old (>70 years) patients with positive lymph nodes and a negative PR status. Lymph node positivity and an S-phase fraction above 5% indicated a worse prognosis in patients who received single radiotherapy; these patients might have benefitted from additional chemotherapy. Adjuvant chemotherapy might also be suitable in patients who are treated with a combination of endocrine therapy and radiotherapy if an S-phase fraction above 5% and a G2–M exceeding rate above 1% are measured, especially in lymph node positive, PR-negative patients with high tumor stages.

Several nomograms for the prediction of overall or disease-free survival of patients with breast cancer have been published (36–39). Different types of parameters were used for their development, but most of them were based on clinicopathologic features only. To our knowledge so far, no nomogram has been developed combining clinicopathologic and genome instability parameters.

As described in Materials and Methods, for the regression model categorized variables were used instead of the continuous measures. We are aware that categorization of those variables means a certain loss of information which possibly decreased the power of the models. Nevertheless, as our main goal was to qualitatively describe which variables can be prognostically useful, we chose to use the categorical approach for an easier interpretation and visualization of the results. Furthermore, not all variables showed a linear relation with the survival time, for example, age. Including those variables as continuous measures would not have provided this information. The cut-off values were chosen on the basis of the hazards of the categories as well as the group size so that robust results could be ensured. Our results provide a basic understanding about which parameters of the genome instability profiles can add prognostic information to standard clinicopathologic factors and should be further evaluated in prospective studies.

To our knowledge, this is one of the largest patient cohort that has ever been analyzed by image cytometry. The high case number allowed us to form clinically relevant subgroups of at least 100 patients to be analyzed separately. All patients were from the greater Stockholm area, which assured standard diagnostics and treatments, and were observed over a postoperative time period of at least 10 years. Nevertheless, besides the strengths of our study, there are weaknesses that have to be mentioned. The patients were diagnosed between 1988 and 2001, therefore data on the HER2 status were not available. Due to the lack of standardization of the image cytometry, the analyzed parameters vary between different studies. For this reason, an independent patient cohort to validate our results was not available and we focused on the explorative data analysis of the subgroups.

In summary, our study showed that in the majority of clinicopathologic and therapeutic breast cancer subgroups, the genome instability profiles can add prognostic information to standard clinicopathologic parameters or even substitute them and serve as single independent predictors. S-phase fraction and SSI were the most meaningful predictors with the highest impact in premenopausal patients. Our results reveal a new possibility for a more precise risk group stratification in patients with breast cancer based on genome instability profiles, which is a further step to individualized medicine. The value of including genome instability data in breast cancer prognostication, especially the assessment of S-phase fraction and SSI, should now be prospectively validated in multicenter studies.

A. Lischka reports personal fees from Ad Infinitum Foundation (PhD scholarship) during the conduct of the study. N. Doberstein reports personal fees from Ad Infinitum Foundation (PhD scholarship) during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

A. Lischka: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing-original draft, writing-review and editing. N. Doberstein: Conceptualization, data curation, formal analysis, investigation, visualization, writing-review and editing. S. Freitag-Wolf: Data curation, formal analysis, investigation, visualization, methodology, writing-original draft, writing-review and editing. A. Koçak: Investigation, writing-review and editing. T. Gemoll: Writing-review and editing. K. Heselmeyer-Haddad: Writing-review and editing. T. Ried: Conceptualization, writing-original draft, project administration, writing-review and editing. G. Auer: Conceptualization, project administration, writing-review and editing. J.K. Habermann: Conceptualization, resources, writing-original draft, project administration, writing-review and editing.

This work was supported in part by the Intramural Research Program of the NCI at the NIH, intramural funding of the University of Lübeck as well as by the Interdisciplinary Center for Biobanking-Lübeck (ICB-L), University of Lübeck. A. Lischka and N. Doberstein obtained a PhD fellowship by the Ad Infinitum Foundation. A. Koçak was supported by the “Deutsche Stiftung für junge Erwachsene mit Krebs”. The authors are grateful to Dr. Stanley Lipkowitz and Dr. Reinhard Ebner for critical comments on the manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bray
F
,
Ferlay
J
,
Soerjomataram
I
,
Siegel
RL
,
Torre
LA
,
Jemal
A
. 
Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J Clin
2018
;
68
:
394
424
.
2.
Senkus
E
,
Kyriakides
S
,
Ohno
S
,
Penault-Llorca
F
,
Poortmans
P
,
Rutgers
E
, et al
Primary breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up
.
Ann Oncol
2015
;
26
:
v8
30
.
3.
Schiavon
G
,
Smith
IE
. 
Status of adjuvant endocrine therapy for breast cancer
.
Breast Cancer Res
2014
;
16
:
206
.
4.
Emens
LA
,
Davidson
NE
. 
Adjuvant hormonal therapy for premenopausal women with breast cancer
.
Clin Cancer Res
2003
;
9
:
486S
94S
.
5.
Early Breast Cancer Trialists' Collaborative Group
. 
Aromatase inhibitors versus tamoxifen in early breast cancer: patient-level meta-analysis of the randomised trials
.
Lancet
2015
;
386
:
1341
52
.
6.
Hudis
CA
. 
Trastuzumab–mechanism of action and use in clinical practice
.
N Engl J Med
2007
;
357
:
39
51
.
7.
Polley
MY
,
Leung
SC
,
McShane
LM
,
Gao
D
,
Hugh
JC
,
Mastropasqua
MG
, et al
An international Ki67 reproducibility study
.
J Natl Cancer Inst
2013
;
105
:
1897
906
.
8.
Paik
S
,
Shak
S
,
Tang
G
,
Kim
C
,
Baker
J
,
Cronin
M
, et al
A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer
.
N Engl J Med
2004
;
351
:
2817
26
.
9.
van 't Veer
LJ
,
Dai
H
,
van de Vijver
MJ
,
He
YD
,
Hart
AA
,
Mao
M
, et al
Gene expression profiling predicts clinical outcome of breast cancer
.
Nature
2002
;
415
:
530
6
.
10.
Ross
JS
. 
Multigene classifiers, prognostic factors, and predictors of breast cancer clinical outcome
.
Adv Anat Pathol
2009
;
16
:
204
15
.
11.
Swanton
C
,
Nicke
B
,
Schuett
M
,
Eklund
AC
,
Ng
C
,
Li
Q
, et al
Chromosomal instability determines taxane response
.
Proc Natl Acad Sci U S A
2009
;
106
:
8671
6
.
12.
Habermann
JK
,
Doering
J
,
Hautaniemi
S
,
Roblick
UJ
,
Bundgen
NK
,
Nicorici
D
, et al
The gene expression signature of genomic instability in breast cancer is an independent predictor of clinical outcome
.
Int J Cancer
2009
;
124
:
1552
64
.
13.
Auer
GU
,
Caspersson
TO
,
Wallgren
AS
. 
DNA content and survival in mammary carcinoma
.
Anal Quant Cytol
1980
;
2
:
161
5
.
14.
Danielsen
HE
,
Pradhan
M
,
Novelli
M
. 
Revisiting tumour aneuploidy - the place of ploidy assessment in the molecular era
.
Nat Rev Clin Oncol
2016
;
13
:
291
304
.
15.
Mandard
AM
,
Denoux
Y
,
Herlin
P
,
Duigou
F
,
van De Vijver
MJ
,
Clahsen
PC
, et al
Prognostic value of DNA cytometry in 281 premenopausal patients with lymph node negative breast carcinoma randomized in a control trial: multivariate analysis with Ki-67 index, mitotic count, and microvessel density
.
Cancer
2000
;
89
:
1748
57
.
16.
Yuan
J
,
Hennessy
C
,
Givan
AL
,
Corbett
IP
,
Henry
JA
,
Sherbet
GV
, et al
Predicting outcome for patients with node negative breast cancer: a comparative study of the value of flow cytometry and cell image analysis for determination of DNA ploidy
.
Br J Cancer
1992
;
65
:
461
5
.
17.
Aubele
M
,
Auer
G
,
Falkmer
U
,
Voss
A
,
Rodenacker
K
,
Rutquist
LE
, et al
Improved prognostication in small (pT1) breast cancers by image cytometry
.
Breast Cancer Res Treat
1995
;
36
:
83
91
.
18.
Aubele
M
,
Auer
G
,
Voss
A
,
Falkmer
U
,
Rutquist
L
,
Hofler
H
. 
Disease-free survival of node-positive breast cancer patients. Improved prognostication by cytometrical parameters
.
Pathol Res Pract
1995
;
191
:
982
90
.
19.
Ermiah
E
,
Abdalla
F
,
Buhmeida
A
,
Alshrad
M
,
Salem
N
,
Pyrhonen
S
, et al
Prognostic significance of DNA image cytometry in Libyan breast cancer
.
Oncology
2012
;
83
:
165
76
.
20.
(UICC) IUAC
.
TNM classification of malignant tumors
.
7th ed
.
Wiley-Blackwell
;
Hoboken, New Jersey, USA
2010
.
21.
Kronenwett
U
,
Huwendiek
S
,
Ostring
C
,
Portwood
N
,
Roblick
UJ
,
Pawitan
Y
, et al
Improved grading of breast adenocarcinomas based on genomic instability
.
Cancer Res
2004
;
64
:
904
9
.
22.
Gaub
J
,
Auer
G
,
Zetterberg
A
. 
Quantitative cytochemical aspects of a combined feulgen-naphthol yellow S staining procedure for the simultaneous determination of nuclear and cytoplasmic proteins and DNA in mammalian cells
.
Exp Cell Res
1975
;
92
:
323
32
.
23.
Cox
DR
. 
Regression models and life-tables
.
J R Stat Soc B
1972
;
34
:
187
.
24.
Bradburn
MJ
,
Clark
TG
,
Love
SB
,
Altman
DG
. 
Survival analysis part II: multivariate data analysis - an introduction to concepts and methods
.
Brit J Cancer
2003
;
89
:
431
6
.
25.
Therneau
TM
. 
A package for survival analysis in S
.
R package version 237-7 2014
. https://cran.r-project.org/web/packages/survival/survival.pdf.
26.
Kast
K
,
Link
T
,
Friedrich
K
,
Petzold
A
,
Niedostatek
A
,
Schoffer
O
, et al
Impact of breast cancer subtypes and patterns of metastasis on outcome
.
Breast Cancer Res Treat
2015
;
150
:
621
9
.
27.
Chen
HL
,
Zhou
MQ
,
Tian
W
,
Meng
KX
,
He
HF
. 
Effect of age on breast cancer patient prognoses: a population-based study using the SEER 18 database
.
PLoS One
2016
;
11
:
e0165409
.
28.
Rudolph
P
,
Kuhling
H
,
Alm
P
,
Ferno
M
,
Baldetorp
B
,
Olsson
H
, et al
Differential prognostic impact of the cyclins E and B in premenopausal and postmenopausal women with lymph node-negative breast cancer
.
Int J Cancer
2003
;
105
:
674
80
.
29.
Auer
G
,
Eriksson
E
,
Azavedo
E
,
Caspersson
T
,
Wallgren
A
. 
Prognostic significance of nuclear DNA content in mammary adenocarcinomas in humans
.
Cancer Res
1984
;
44
:
394
6
.
30.
Spyratos
F
,
Briffod
M
. 
DNA ploidy and S-phase fraction by image and flow cytometry in breast cancer fine-needle cytopunctures
.
Mod Pathol
1997
;
10
:
556
63
.
31.
Dodd
LG
,
Kerns
BJ
,
Dodge
RK
,
Layfield
LJ
. 
Intratumoral heterogeneity in primary breast carcinoma: study of concurrent parameters
.
J Surg Oncol
1997
;
64
:
280
7
.
32.
Morris
LG
,
Riaz
N
,
Desrichard
A
,
Senbabaoglu
Y
,
Hakimi
AA
,
Makarov
V
, et al
Pan-cancer analysis of intratumor heterogeneity as a prognostic determinant of survival
.
Oncotarget
2016
;
7
:
10051
63
.
33.
Ermiah
E
,
Buhmeida
A
,
Abdalla
F
,
Khaled
BR
,
Salem
N
,
Pyrhonen
S
, et al
Prognostic value of proliferation markers: immunohistochemical ki-67 expression and cytometric s-phase fraction of women with breast cancer in libya
.
J Cancer
2012
;
3
:
421
31
.
34.
Martinez-Arribas
F
,
Nunez
MJ
,
Piqueras
V
,
Lucas
AR
,
Sanchez
J
,
Tejerina
A
, et al
Flow cytometry vs. Ki67 labelling index in breast cancer: a prospective evaluation of 181 cases
.
Anticancer Res
2002
;
22
:
295
8
.
35.
Spyratos
F
,
Ferrero-Pous
M
,
Trassard
M
,
Hacene
K
,
Phillips
E
,
Tubiana-Hulin
M
, et al
Correlation between MIB-1 and other proliferation markers: clinical implications of the MIB-1 cutoff value
.
Cancer
2002
;
94
:
2151
9
.
36.
Su
J
,
Miao
LF
,
Ye
XH
,
Cui
MS
,
He
XF
. 
Development of prognostic signature and nomogram for patients with breast cancer
.
Medicine
2019
;
98
:
e14617
.
37.
Li
S
,
Zhao
J
,
Zhu
L
,
Su
F
,
Chen
K
. 
Development and validation of a nomogram predicting the overall survival of stage IV breast cancer patients
.
Cancer Med
2017
;
6
:
2586
94
.
38.
Gong
Y
,
Ji
P
,
Sun
W
,
Jiang
YZ
,
Hu
X
,
Shao
ZM
. 
Development and validation of nomograms for predicting overall and breast cancer-specific survival in young women with breast cancer: a population-based study
.
Transl Oncol
2018
;
11
:
1334
42
.
39.
Rudloff
U
,
Jacks
LM
,
Goldberg
JI
,
Wynveen
CA
,
Brogi
E
,
Patil
S
, et al
Nomogram for predicting the risk of local recurrence after breast-conserving surgery for ductal carcinoma in situ
.
J Clin Oncol
2010
;
28
:
3762
9
.