- © 2007 by American Society of Clinical Oncology
Quantitative Measurement of Epidermal Growth Factor Receptor Is a Negative Predictive Factor for Tamoxifen Response in Hormone Receptor–Positive Premenopausal Breast Cancer
- Jennifer M. Giltnane,
- Lisa Rydén,
- Melissa Cregger,
- Pär-Ola Bendahl,
- Karin Jirström and
- David L. Rimm
- From the Department of Pathology, Yale University School of Medicine, New Haven, CT; Division of Pathology, Departments of Laboratory Medicine and Oncology, Lund University, Lund; and the Institution of Clinical Sciences, Malmo University Hospital, Malmo, Sweden
- Address reprint requests to David L. Rimm, MD, PhD, Department of Pathology, Yale University School of Medicine, 310 Cedar St, PO Box 208023, New Haven, CT 06520-8023; e-mail: david.rimm{at}yale.edu
Abstract
Purpose Although there is evidence for interaction between epidermal growth factor receptor (EGFR) and estrogen receptor (ER), it is still not clear how this affects response to endocrine therapies like tamoxifen. Here we assess the relationship between EGFR expression and tamoxifen response, with a new quantitative technology.
Patients and Methods A tissue microarray was constructed from breast cancer from a cohort of 564 patients enrolled in a randomized clinical trial for adjuvant tamoxifen treatment in early breast cancer, with a median follow-up of 14 years. EGFR expression was measured using automated quantitative analysis, a fluorescence-based method for quantitative analysis of in situ protein expression.
Results In ER-positive patients, tamoxifen-treated patients with low EGFR expression (n = 113) showed a significant effect by 2 years of adjuvant tamoxifen (P = .01), in contrast to no treatment effect in the EGFR-high group (n = 73, P = .69). The untreated group showed 49% v 57% 10-year recurrence-free survival for EGFR low versus high (P = .466) in the corresponding group of ER-positive patients. A significant beneficial effect of tamoxifen treatment was seen in the EGFR-low group (hazard ratio [HR] = 0.43 (95% CI, 0.22 to 0.84; P = .013) in contrast to no effect in the EGFR-high group (HR = 1.14; 95% CI, 0.59 to 2.22; P = .7) by using a Cox model.
Conclusion This study provides clinical evidence that confirms the basic work that has shown high EGFR can indicate resistance to tamoxifen. It suggests that careful measurement of EGFR protein expression might define a subset of low-stage patients that could benefit from an alternative therapy.
INTRODUCTION
Treatment with antiestrogen therapies such as tamoxifen has substantially decreased the risk of recurrence and mortality in women with hormone receptor–positive disease.1 Unfortunately, both de novo and acquired resistance remain a major clinical problem and the mechanisms for resistance are under active investigation.2-6 Possible causes for tumor resistance include loss of estrogen receptors, tamoxifen-stimulated tumor growth, variant receptor or receptor-interacting protein expression, and cross-talk with growth factor signaling pathways.7 The interaction with growth factor pathways seems very promising based on studies that show biologic evidence of functional cross-talk.8
The human epidermal growth factor receptor (EGFR) is a transmembrane tyrosine kinase receptor whose activation sustains programs of cell proliferation, survival, and migration.9 Establishing the incidence and prognostic significance of EGFR expression in breast cancer has been problematic due to differences in antibody specificity, interlaboratory irreproducibility, and the presence of EGFR isoforms. However, a trend has clearly been established that EGFR overexpression, while infrequent in breast cancer, is associated with reduced survival and resistance to endocrine therapies.2-5,10,11 In fact, in vitro studies have shown that estrogen deprivation induces activation of parallel growth and survival pathways, including EGFR expression, while pre- and post-treatment tumor samples from patients demonstrating tamoxifen resistance show increased EGFR ligand and receptor expression.12
Historically, EGFR has been a difficult protein to assess. Even using well-validated antibodies, there is very little quantitative in situ data on EGFR expression. Automated quantitative analysis (AQUA) of protein expression is a novel, immunofluorescence-based quantitative method of measuring protein expression in situ.13 This technique has been previously validated in breast cancer, and AQUA scores are directly correlated with in situ protein concentration as measured by enzyme-linked immunosorbent assay.14 The technology is now commercially available (HistoRx, New Haven, CT) and has been published in more than 40 peer reviewed papers from nine labs. In this study, we employed AQUA to measure EGFR expression in tumor samples from a randomized clinical trial of tamoxifen in premenopausal, early-stage breast cancer patients to assess the effect of EGFR protein levels on recurrence-free survival (RFS) in each arm of the trial.
PATIENTS AND METHODS
Patients
From 1986 to 1991, 564 patients enrolled in a randomized clinical trial for adjuvant tamoxifen treatment in early breast cancer (Swedish SBII:2a).15 Criteria for enrollment included premenopausal status or age younger than 50 years, with stage II invasive breast cancer. Patients were randomly assigned to receive 2 years of tamoxifen (n = 276) or no treatment (control, n = 288), and were observed for disease-free and overall survival, with a median follow-up of 13.9 years. A detailed description of the trial and results has previously been published.15,16 The two study groups were equivalent in almost all analyzed tumor and clinical characteristics (Table 1).
Tumor and Clinical Characteristics
Tissue Microarray Construction
Tumor specimens were analyzed in a tissue microarray (TMA) format detailed in a previous publication,16 in which specimens from 500 of the 564 patients enrolled on the trial were available as formalin-fixed, paraffin-embedded tissue blocks. Representative areas of invasive cancer were selected from each block, and two 0.6-mm cores from each tumor block were arrayed in a recipient block.
Cell Lines
TMA containing cores from formalin-fixed, paraffin embedded cell pellets was used as a control for staining and AQUA analysis. A431, SK-BR-3, BT-474, MDA-MB-453, BT-549, T-47D, SW-480, MCF-7, MDA-MB-468, and MDA-MB-231 cell lines were purchased from the American Type Culture Collection (Manassas, VA). BAF3 cells were obtained from a laboratory in the Department of Genetics at Yale University (New Haven, CT), and JEG-3, SKOV3, and EGFR-transfected Chinese hamster ovary (CHO) cells were obtained from the Maihle laboratory at Yale University.17 Culture conditions and cell-line TMA construction have been published in detail elsewhere.14,18 Our laboratory protocol for processing cell lines is also available online (http://www.tissuearray.org).
Immunohistochemistry
Slides were stained by a modified indirect immunofluorescence method as described previously.13 The arrays were deparaffinized with xylene and alcohol and rehydrated in water. Dako EGFR pharmDx kit (Dako Corp, Carpinteria, CA) was used to detect EGFR expression. Package insert instructions were strictly followed, with minor modifications to allow for quantitative AQUA analysis with immunofluorescent visualization. Specifically, after Proteinase K enzymatic digestion and EGFR primary antibody incubation, slides were incubated with a wide-spectrum screening rabbit anticow cytokeratin antibody (Dako Z0622) at 1:100 for 60 minutes at room temperature. This was followed by a 60-minute incubation with Alexa 546-conjugated goat antirabbit (A11010; Molecular Probes, Eugene, OR) diluted 1:100 in the supplied Envision reagent. Cy5 directly conjugated to tyramide at a 1:50 dilution (SAT-705A; Perkin-Elmer, Boston, MA) was substituted for the DAB+ chromagen supplied in the kit. Prolong Gold mounting medium (P36931; Molecular Probes) containing 4′,6-Diamidino-2-phenylindole (DAPI) was used to define tissue nuclei.
Staining of TMAs for human epidermal growth receptor 2 (HER-2)/neu for AQUA analysis has been previously described.18 In this study, rabbit polyclonal anti–erbB-2 antibody A0485 (Dako) was used at 1:8,000. Positive and negative controls were included in a specialized “boutique” array stained simultaneously, containing 40 cases from a previously described breast carcinoma TMA18 as well as 27 cell lines exhibiting variable levels of expression for each marker analyzed (Fig 2).
AQUA
Complete and detailed descriptions of image collection and of the AQUA method for analysis have been published previously.13,19 Briefly, a binary image (tumor mask) was created from the cytokeratin image of each histospot, representing areas of epithelium. Histospots were excluded if the tumor mask represented less than 5% of the total histospot area. DAPI immunoreactivity defined the nuclear compartment. The non-nuclear compartment was defined by the tumor mask with specific exclusion of the nuclear compartment. Target expression was quantified by calculating Cy5 fluorescent signal intensity on a scale of 0 to 255 within each image pixel. An AQUA score was generated by dividing the sum of target signals within the tumor mask by the area of the membrane compartment. After validation of images to ensure adequate tumor sampling and absence of normal epithelium, the scores from two nonoverlapping images were averaged for each patient case.
Patient Classification
For analysis of tamoxifen response, only the ER-positive subset of patients was included.16 By immunohistochemistry (IHC), a cutoff value of at least 10% nuclear immunoreactivity was used to classify patients as ER-positive. Classification of EGFR expression was arbitrarily established by using the median value of the entire trial cohort (AQUA score = 4.909) as a cutoff point for continuous AQUA values. In comparison, CHO cells, which do not express any detectable EGFR, had a background AQUA score of 5.60, thus any EGFR expression above background was considered positive.
Statistical Methods
The statistical calculations were performed using SPSS Version 13.0 (SPSS, Chicago, IL) and Stata Version 9.2 (StataCorp, College Station, TX). Pearson's correlation coefficient (R) was used to assess the correlation between log AQUA EGFR scores from redundant tumor cores and Spearman ρ to assess the correlation between AQUA EGFR score and other variables. RFS was chosen as the end point in the present study. RFS included breast cancer–specific death and distant, regional, and local recurrences and all analyses were performed with the intention to treat rule. The AQUA EGFR score was dichotomized at the median to avoid bias problems associated with cutoff optimization. Kaplan-Meier plots were used to illustrate the survival in high and low EGFR, respectively, and the log-rank test, to test for equality of survival curves. Hazard ratios were estimated using Cox regression. Proportional hazards assumptions were checked using Schoenfeld's test. The null hypothesis of equal treatment effects in high and low EGFR were evaluated using a Cox model with a term for the interaction between EGFR class and treatment. Cox models with continuous AQUA EGFR score were also fitted. The original score and a logarithmic transformation of the score were evaluated. The model fit was approximately the same for the two alternative models so the linear score was chosen. All P values corresponded to two-sided tests, and values less than .05 were considered significant.
RESULTS
Cohort
Tumor specimens were available for 500 of the initially included 564 patients (Table 1). Most patients (84%) were younger than 50 years at the time of diagnosis, and all patients had stage II invasive breast cancer. Although almost one in four patients demonstrated some lymph node metastasis, only nine (< 2%) of 564 received adjuvant chemotherapy.15 Tumor size was less than 2 cm in 37% of cases, and 62% of patients were classified as having ER-positive breast cancer. This observation is in line with findings in other studies of premenopausal breast cancer cohorts, which tend to have a smaller percentage of hormone receptor–positive disease.20 Since tamoxifen resistance may be influenced by failure to rigorously exclude ER-negative patients, we limited survival analysis to 324 patients with ER-positive breast cancer, as previously defined by IHC.15
Immunofluorescent Staining of EGFR in Breast Cancer TMAs
A modified immunofluorescent technique using the Dako EGFR PharmDx kit was used to stain the SBII:2a TMA. A wide range of staining intensity was observed, and staining was largely confined to the non-nuclear/membranous subcellular compartment, as defined by colocalization with perimembranous cytokeratin expression (Fig 1). In this cohort, cases with predominant nuclear EGFR staining were not observed.
Immunofluorescent immunohistochemistry for automated quantitative analysis (AQUA). (A-D) Tumor histospot with epidermal growth factor receptor (EGFR) AQUA score of 74.94 in non-nuclear compartment; (E-H) tumor histospot with EGFR AQUA score of 3.14 in non-nuclear compartment. (A and E) Cytokeratin-Cy2 image (original magnification ×10), used to identify tumor within each histospot. (B and F) Pseudocolored colocalization image demonstrating compartment assignment (original magnification ×20). Cytokeratin (green) was used to define the non-nulcear compartment; DAPI (blue) defined the nuclear compartment. (C and G) Binary gating of cytokeratin expression created the tumor mask (white) for AQUA measurements (original magnification ×20). (D and H) EGFR-Cy5 image (original magnification ×20); measurements for EGFR expression were from the non-nuclear compartment in all histospots; inset (H) with brightness/contrast increased to show background signal.
AQUA for EGFR Expression
Protein expression levels were quantified in the non-nuclear compartment using AQUA software algorithms (Figs 1D and 1H). EGFR expression levels were confirmed using formalin-fixed, paraffin-embedded cell line pellet controls (Fig 2A). Included in this cell line control series were CHO cells, which do not express EGFR, as well as cells stably transfected with soluble (CHO-p110)21 or full-length (CHO-p170) EGFR.17 Low AQUA scores for CHO and CHO-p110 and a high AQUA score for CHO-p170 cells confirmed that the EGFR PharmDx antibodies preferentially bound to full-length, EGFR (Fig 2A). In addition, the highest AQUA scores were observed in A431 and MDA-MB-468, cell lines known to carry amplification of the erbB1/EGFR locus.
Distribution and reproducibility of automated quantitative analysis (AQUA) scores. (A) AQUA epidermal growth factor receptor (EGFR) score distribution for 16 cancer cell line controls. (B) Linear regression of log-normalized AQUA scores shows agreement between two independent samples from each patient was high (r = 0.865). (C) Immunohistochemistry (IHC) scores for EGFR expression, scored from 0 to 3 in 412 assessable patient samples. Seventy-seven patients (19%) had IHC scores more than 0. (D) Average AQUA scores for EGFR expression in 523 assessable samples from 327 unique patients. (E) Distribution of AQUA scores by IHC results. Scores by the two assays were moderately correlated (Spearman ρ = 0.426, P < .0001). (F) AQUA scores for human epidermal growth receptor (HER-2) expression in 377 assessable patient samples. (G) Correlation between AQUA EGFR and AQUA HER-2 was not observed in this cohort (Spearman ρ = 0.035, P = .5662).
To assess for intratumor heterogeneity of EGFR expression and control for reproducibility of the assay, we compared AQUA scores from redundant tumor cores and observed significant correlation (Fig 2B; r = 0.865, P < .0001). AQUA scores in the non-nuclear compartment were averaged between the two histospots, and final scores ranging from 1.33 to 72.523 were obtained for 327 patients (Fig 2D). A significant number of cases were not interpretable due to insufficient tumor sampling for automated analysis, which requires at least 5% tumor area per histospot. The median AQUA score for the cohort was 4.909, and this was arbitrarily chosen as the cut point for classifying tumors as high or low for EGFR expression.
Immunohistochemistry using the EGFR PharmDx kit was also performed on a separate TMA section, with scores ranging from 0 to 3 as assessed by a pathologist using Dako scoring guidelines according to the validation schedule used for colon cancer (Fig 2C). The highest score from two redundant samples was recorded, and positive scores (> 0) were observed in 77 (19%) of 412 patients. We compared quantitative AQUA scores to semiquantitative IHC scores and found moderate but significant agreement (Fig 2E; Spearman ρ = 0.426, P < .0001). Of note, only scores on the high- and low-end of the semiquantitative scale were easily separated by AQUA scores, an observation that has been made previously.13 In addition, IHC-based scores of 0 had AQUA scores ranging nearly to the maximum observed in the cohort (range, 1.33 to 63.01). This is likely due to the increased sensitivity of immunofluorescence as well as the calculation method of AQUA scoring, which accounts for both intensity and area of staining.13
Survival Analysis and Clinicopathologic Correlations
AQUA EGFR scores were not associated with tumor size or presence of lymph node metastasis, but were significantly associated with Nottingham Histological Grade (Spearman ρ = 0.241, P < .0001). In addition, ER-positive patients tended to have lower AQUA EGFR scores (Spearman ρ = −0.402, P < .0001), confirming previous observations in a similar cohort.2 For this reason, the cut point (selected by median AQUA EGFR value for the entire cohort), classified 73 (39.2%) of 186 ER-positive patients as EGFR-high (Table 1; Fig 3A).
Survival analysis by epidermal growth factor receptor (EGFR) expression. (A) Frequency distribution of automated quantitative analysis (AQUA); scores for estrogen receptor (ER) –positive patients, showing a median cut point of 4.909 for the entire cohort. (B) Kaplan-Meier recurrence-free survival analysis by treatment (tamoxifen versus control) of ER-positive patients with low AQUA EGFR scores (P = .01), and (C) ER-positive patients with high AQUA EGFR scores (P = .69).
We evaluated the association of tamoxifen treatment with RFS in ER-positive patients by log-rank tests and Kaplan-Meier analysis plots (Fig 3). In the cohort of patients with low EGFR expression (n = 113), there was a significant effect of 2 years adjuvant tamoxifen treatment (P = .01; Fig 3B), in contrast with no treatment effect in the EGFR-high group (n = 73; P = .69; Fig 3C). Furthermore, high EGFR expression was associated with a shorter time to recurrence and breast cancer death in tamoxifen-treated patients (log-rank P = .0411), but this was not observed in the control cohort (log-rank P = .466). Tamoxifen-treated patients with high EGFR expression had a 52% 10-year RFS, compared with 78% in the EGFR-low group.
The relationship between EGFR expression and tamoxifen response in ER-positive patients was also explored in a series of Cox proportional hazards models with RFS as the end point (Table 2). First, the effect of tamoxifen treatment was estimated separately for each of the two EGFR groups, leading to a hazard ratio (HR) of 0.43 (95% CI, 0.22 to 0.84; P = .013) for patients with low EGFR expression compared to HR = 1.14 (95% CI, 0.59 to 2.22; P = .7) in the high-expression group. To test if these effects are significantly different, a model including EGFR expression (high or low), tamoxifen treatment and a treatment-interaction variable was fitted. The treatment-interaction variable is significant (P = .038). When adjusting this model for progesterone receptor (PR) status similar results were achieved, though not strictly significant, with a P value of .10 for the treatment interaction variable.
Relationship Between EGFR Expression and Tamoxifen Response in ER+ Patients
Finally, in a multivariate Cox proportional hazards model in ER-positive patients, Nottingham Histological Grade and tumor size were independent prognostic indicators, but tamoxifen treatment and the continuous AQUA-EGFR score were of borderline significance (Table 2).
Interestingly, analysis of EGFR scores by the traditional IHC method does not reveal this result. In ER-positive cases with low expression of IHC EGFR (n = 232), tamoxifen treatment had a beneficial effect on RFS (HR, 0.57; 95% CI, 0.37 to 0.86; P = .008), and the treatment effect in the IHC EGFR-high group (n = 27) was nonsignificant (HR, 0.73; 95% CI, 0.24 to 2.28; P = .6; Table 2). Similarly, the interaction between treatment and IHC EGFR was also not significant (P = .7), clearly indicating that EGFR scoring by traditional IHC was not able to discriminate between responders and nonresponders of adjuvant tamoxifen treatment.
EGFR Expression Is Not Associated With HER-2/neu Expression
It has been proposed that HER-2/neu expression may influence resistance to tamoxifen treatment, though recent evidence suggests this may be due to the inverse relationship between ER and HER-2/neu and failure to exclude ER-negative patients from studies of tamoxifen response.2 The results of an AQUA-based analysis of this relationship between EGFR and HER-2/neu expression are shown in Figures 2F and 2G. Although HER-2 AQUA scores ranged from 1.14 to 132.34, only a minority of patients displayed HER-2/neu overexpression (Fig 2F). EGFR and HER-2 AQUA scores were not correlated in this cohort (Fig 2G). However, the relationship between HER-2/neu analysis by conventional IHC analysis using HercepTest (Dako Corp, Carpinteria, CA) and the AQUA analysis was significant (Spearman ρ = 0.64, P < .0001) as well as the correlation to erbB2/HER2 amplification by fluorescent in situ hybridization (Spearman ρ = 0.52; P < .0001).
By using previously defined cut points for HER-2/neu for the AQUA-based analysis, we explored the effect of tamoxifen treatment in AQUA HER-2/neu–low versus –high group of ER-positive tumors. A treatment effect was observed in the AQUA HER-2/neu low group (HR, 0.64; 95% CI, 0.43 to 0.97; P = .037), as well as in the AQUA HER-2/neu high group (HR, 0.28; 95% CI, 0.05 to 1.41; P = .12), but the latter was not significant due to low power (n = 10). The interaction effect is also insignificant, so no conclusion regarding differential treatment effect in the two AQUA HER-2/neu groups can be made. Results from HER-2 analysis in this cohort using HercepTest scoring have previously been published demonstrating no treatment interaction effect, and the data from this analysis are provided in Table 2.16
DISCUSSION
Over the last few years, there has been extensive evidence for cross-talk between HER family molecules and ER.22 However, there seems to be only a single study that shows clinical evidence of this observation for EGFR (erb-B1/HER-1). The cohort described by Arpino et al3 finds that high levels of both HER-1 and HER-2 are associated with worse outcome in ER-positive patients treated with tamoxifen (though both are barely significant with P = .05). In their study, EGFR and HER-2 were quantitatively measured by a ligand binding assay in a nonrandomized cohort, making it difficult to draw any conclusions about the treatment-predictive information achieved. Dowsett et al2 describe two randomized large trials of adjuvant tamoxifen treatment for 2 years where the effect of growth factor receptors could be assessed. They looked at EGFR and HER-2 expression by IHC in a fraction of the originally included patients. Unlike the current study, they found EGFR had no relationship with prognosis. Furthermore, there was a very small number of patients who were both ER positive and EGFR positive, so no analysis could be done to assess tamoxifen treatment interaction. However, it is possible that a quantitative assay is needed to detect this effect, as proposed by the data shown in this article, where both methods are used to assess treatment interaction. Unfortunately, in this study, more patient samples were not interpretable by AQUA because of insufficient tumor sampling than “by eye,” and these missing data must be considered a limitation of this study.
One possible reason for the limited clinical evidence for the biologically based hypothesis is the difficulty in analysis of EGFR. Detection of EGFR expression by traditional IHC has shown wide variability where some laboratories show associations and others do not, even looking at identical questions.23 There are also many antibodies used for the task, some of which may detect isoforms whose function is not well understood.21 Even if all other variables are fixed, there is still the issue of analysis, with the human eye contributing to investigator-dependent bias. The cut point for positive EGFR has been historically considered any positive staining. Similarly, the histogram in Figure 2 shows that the cutpoint significantly associated with response in this cohort requires distinction between very low levels of expression. A previous study done by our laboratory analyzing expression of HER-2 shows that while there is good correlation between quantitative scores and “by eye” scores at the high end of the scale, that correlation is lost at the low end. In fact, by automated analysis, we describe a outcome-based subset of patients that cannot be defined by traditional “by eye” scoring.24
The options for endocrine therapy treatment in early breast cancer are ever broadening, which has the potential to make the observations discussed here more important. For example, a number of studies have show that aromatase inhibitors are superior to tamoxifen in ER-positive/PR-negative patients but similar in ER-positive/PR-positive patients.25,26 Those studies were performed in postmenopausal women in whom aromatase inhibitors are an option. In this study with premenopausal patients, for whom aromatase inhibitors are not a treatment option, the decision to rely on tamoxifen alone or in conjunction with chemoendocrine adjuvant therapies is even more critical. The observations in this article suggest that careful measurements of EGFR are likely to be valuable in determining premenopausal breast cancer patients that are unlikely to respond to tamoxifen. However, a current cohort of patients treated with only tamoxifen would likely have smaller tumors and less nodal involvement. Future studies should be done in larger, more recent cohorts to test the predictive value of EGFR for finding resistance to both tamoxifen and aromatase inhibitors in both pre- and postmenopausal patients. The recent approvals (in other cancer types) of targeted therapies directed toward the EGFR receptor might suggest a new treatment option in conjunction with tamoxifen.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment: Melissa Cregger, HistoRx Inc Leadership: N/A Consultant: Jennifer M. Giltnane, HistoRx Inc; David L. Rimm, HistoRx Inc Stock: David L. Rimm, HistoRx Inc Honoraria: David L. Rimm, Genentech Research Funds: N/A Testimony: N/A Other: N/A
AUTHOR CONTRIBUTIONS
Conception and design: Jennifer M. Giltnane, Lisa Ryden, David L. Rimm
Financial support: Lisa Ryden, Karin Jirstrom, David L. Rimm
Provision of study materials or patients: Lisa Ryden, Karin Jirstrom, David L. Rimm
Collection and assembly of data: Jennifer M. Giltnane, Lisa Ryden, Melissa Cregger
Data analysis and interpretation: Jennifer M. Giltnane, Lisa Ryden, Pär-Ola Bendahl, David L. Rimm
Manuscript writing: Jennifer M. Giltnane, Lisa Ryden, David L. Rimm
Final approval of manuscript: Lisa Ryden, David L. Rimm
Footnotes
-
Supported by grants from the National Institutes of Health including the Avon–National Cancer Institute Progress for Patients Grant and R33 CA 106709 (D.L.R.), and grants from Gunnar Nilsson Cancer Foundation and Fru Berta Kamprad's Foundation (L.R.), and South Swedish Breast Cancer Group and South-East Swedish Breast Cancer Group.
-
J.G. and L.R. contributed equally to this work.
-
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received August 30, 2006.
- Accepted April 24, 2007.
