Dataset on growth factor levels and insulin use in patients with diabetes mellitus and incident breast cancer

Growth factor profiles could be influenced by the utilization of exogenous insulin. The data presented shows the relationship between pre-existing use of injectable insulin in women diagnosed with breast cancer and type 2 diabetes mellitus, the growth factor profiles at the time of breast cancer diagnosis, and subsequent cancer outcomes. A Pearson correlation analysis evaluating the relationship between growth factors stratified by of insulin use and controls is also provided.

Reported data has the potential to guide future research evaluating insulin-induced growth factor modulation in breast cancer.
Our observations may assist future studies in evaluating the relationship between insulin safety and effectiveness and growth factors production in cancer.

Data
Reported data represents the observed association between use of injectable insulin preceding breast cancer and the growth factor profiles at the time of cancer diagnosis in women with diabetes mellitus (Table 1). Data in Table 2 includes the observed correlations between growth factors stratified by type 2 diabetes mellitus pharmacotherapy and controls. C-peptide correlation with each of the studied growth factors is presented in Table 2, however details regarding its determination from plasma, association with cancer outcomes and use of injectable insulin has been previously reported by us [1]. * Overall survival (OS)-and disease-free survival (DFS)-optimized growth factor ranges associated with poorer outcomes are represented in bold. BLQ ¼below limit of quantitation. p 1 ¼ pairwise comparison of controls with the no insulin group, p 2 ¼ pairwise comparison of controls with the any insulin group, and p 3 ¼pairwise comparison of the no insulin and any insulin groups. Global Test ¼ significance test across all groups. MVP ¼p-value of the multivariate adjusted analysis. Epidermal growth factor (EGF), fibroblast Growth Factor 2 (FGF-2), hepatocyte growth factor (HGF), platelet-derived growth factor BB (PDGF-BB), tumor growth factor (TGF), vascular endothelial growth factor (VEGF).

Experimental design, materials and methods
Evaluation of growth factor profile association with injectable insulin use and BC outcomes was carried out under two protocols approved by both Roswell Park Cancer Institute (EDR154409 and NHR009010) and the State University of New York at Buffalo (PHP0840409E). Demographic and clinical patient information was linked with cancer outcomes and growth factor profiles of Table 2 Growth factor correlations by insulin use. corresponding plasma specimen harvested at BC diagnosis and banked in the Roswell Park Cancer Institute Data Bank and Bio-Repository.

Study population
All incident breast cancer cases diagnosed at Roswell Park Cancer Institute (01/01/2003 À 12/31/ 2009) were considered for inclusion (n ¼2194). Medical and pharmacotherapy history were used to determine the baseline presence of diabetes.

Inclusion and exclusion criteria
All adult women with pre-existing diabetes at breast cancer diagnosis having available banked treatment-naïve plasma specimens (blood collected prior to initiation of any cancer-related therapysurgery, radiation or pharmacotherapy) in the Institute's Data Bank and Bio-Repository were included.
Subjects were excluded if they had prior cancer history or unclear date of diagnosis, incomplete clinical records, type 1 or unclear diabetes status. For a specific breakdown of excluded subjects, please see the original research article by Wintrob et al. [1].
A total of 97 female subjects with breast cancer and baseline diabetes mellitus were eligible for inclusion in this analysis.

Control-matching approach
Each of the 97 adult female subjects with breast cancer and diabetes mellitus (defined as "cases") was matched with two other female subjects diagnosed with breast cancer, but without baseline diabetes mellitus (defined as "controls"). The following matching criteria were used: age at diagnosis, body mass index category, ethnicity, menopausal status and tumor stage (as per the American Joint Committee on Cancer). Some matching limitations applied [1].

Demographic and clinical data collection
Clinical and treatment history was documented as previously described [1]. Vital status was obtained from the Institute's Tumor Registry, a database updated biannually with data obtained from the National Comprehensive Cancer Networks' Oncology Outcomes Database. Outcomes of interest were breast cancer recurrence and/or death.

Plasma specimen storage and retrieval
All the plasma specimens retrieved from long-term storage were individually aliquoted in color coded vials labeled with unique, subject specific barcodes. Overall duration of freezing time was accounted for all matched controls ensuring that the case and matched control specimens had similar overall storage conditions. Only two instances of freeze-thaw were allowed between biobank retrieval and biomarker analyses: aliquoting procedure step and actual assay.

Luminex s assays
A total of 6 biomarkers (epidermal growth factor, fibroblast growth factor 2, vascular endothelial growth factor, hepatocyte growth factor, platelet-derived growth factor BB, and tumor growth factorβ) were quantified according to the manufacturer protocol. The following Luminex s biomarker panels were utilized in this study: TGFB-64K (tumor growth factor-β), HCYTOMAG-60K (platelet-derived growth factor BB), and HAGP1MAG-12K (epidermal growth factor, fibroblast growth factor 2, vascular endothelial growth factor, and hepatocyte growth factor) produced by Millipore Corporation, Billerica, MA. C-peptide determinations were done according to the manufacturer protocol as previously reported [2].

Biomarker-pharmacotherapy association analysis
Biomarker cut-point optimization was performed for each analyzed biomarker. Biomarker levels constituted the continuous independent variable that was subdivided into two groups that optimized the log rank test among all possible cut-point selections yielding a minimum of 10 patients in any resulting group. Quartiles were also constructed. The resultant biomarker categories were then tested for association with type 2 diabetes mellitus therapy and controls by Fisher's exact test. The continuous biomarker levels were also tested for association with diabetes therapy and controls across groups by the Kruskal-Wallis test and pairwise by the Wilcoxon rank sum. Multivariate adjustments were performed accounting for age, tumor stage, body mass index, estrogen receptor status, and cumulative comorbidity. The biomarker analysis was performed using R Version 2.15.3. Please see the original article for an illustration of the analysis workflow [1].
Correlations between biomarkers stratified by type 2 diabetes mellitus pharmacotherapy and controls were assessed by the Pearson method. Correlation models were constructed both with and without adjustment for age, body mass index, and the combined comorbidity index. Correlation analyses were performed using SAS Version 9.4.