Browse the corpus

INTRODUCTION Breast cancer has been recognized as a complex, heterogeneous disease, encompassing multiple cell populations with different risk factors, histological features, clinical behaviors and responses to therapy[1-3]. While breast cancer diagnosis was initially based on tumor size, histological classification systems were later developed to categorize breast tumors into subgroups. Primary breast carcinomas are first classified as either in situ or invasive tumors. Ductal carcinoma in situ (DCIS) of the breast is an early stage, non-invasive breast tumor commonly diagnosed by mammography screening. DCIS accounts for 15%-20% of all newly diagnosed breast cancer cases, and is characterized by the intraductal proliferation of malignant epithelial cells without invasion through the basement membrane into the surrounding tissue[4]. In contrast, invasive breast carcinoma is able to invade through the basement membrane into the surrounding stroma. Invasive breast carcinomas have been extensively studied and found to be composed of numerous heterogeneous histological subtypes[5]. Similar to invasive breast carcinoma, divergent histological types of DCIS lesions have been recognized, and various classification systems have been developed to characterize DCIS tumors[6]. Approximately 15%-30% of DCIS patients relapse within 10 years after surgical lumpectomy[7], and it is of urgent clinical need to have an effective classification system to identify DCIS with a high-risk of tumor recurrence. Owing to subjective interpretation of lesion morphology by pathologists, inconsistency in DCIS classification cannot be avoided[8]. Histological classification of heterogeneous DCIS lesions is not sufficient to identify molecularly heterogeneous DCIS subgroups, and additional classification approaches are necessary for the pathological characterization of DCIS and identification of more effective therapeutic options.

ification cannot be avoided[8]. Histological classification of heterogeneous DCIS lesions is not sufficient to identify molecularly heterogeneous DCIS subgroups, and additional classification approaches are necessary for the pathological characterization of DCIS and identification of more effective therapeutic options. Gene expression profiling has emerged as a useful system for breast cancer classification[9-11], and has been used to define five intrinsic molecular subtypes of invasive breast carcinoma: Luminal A, luminal B, HER2-enriched, normal- and basal-like[9,10]. Following this discovery, additional subgroups of breast cancer were identified, including the interferon-enriched[12], molecular apocrine[13] and claudin-low subgroups[14]. Given that these subtypes possess different molecular alterations, they display distinct clinical outcomes and therapeutic responses. The basal-like subtype is highly aggressive and therefore of particular clinical relevance. Basal-like breast cancers are more likely to occur in younger, African American women, and are associated with breast cancer susceptibility (BRCA) gene mutations. They are characterized by high tumor grade, proliferation rate, frequency of recurrence, and the presence of p53 mutations. Patients with basal-like breast cancers frequently have poor prognosis, and are difficult to treat due to the lack of effective targeted therapies[10,11,15]. Breast tumors categorized as basal-like display gene expression signatures similar to normal basal/myoepithelial breast cells (myofibroblast-like breast epithelial cells located between breast ductal epithelial cells and the basement membrane), including high-molecular weight basal cytokeratins (CK5/6, CK14 and CK17)[16].

ast tumors categorized as basal-like display gene expression signatures similar to normal basal/myoepithelial breast cells (myofibroblast-like breast epithelial cells located between breast ductal epithelial cells and the basement membrane), including high-molecular weight basal cytokeratins (CK5/6, CK14 and CK17)[16]. The majority of diagnosed basal-like breast cancer cases are triple-negative breast cancers (TNBC), which lack expression of hormone receptors [estrogen receptor (ER) and progesterone receptor (PR)] and overexpression/amplification of HER2[9,10,15]. Although there is significant overlap between basal-like breast cancers and TNBC, they are not identical. Approximately 70%-80% of basal-like breast cancers have been identified as triple-negative, basal-like breast cancer (TN-BLBC)[15,17,18]. The remaining non-triple-negative basal-like breast cancers share similar gene expression profiles with TN-BLBC, but might have gained additional genetic and/or epigenetic aberrations due to increased genomic instability[17,18]. Using gene expression profiling analysis, Lehmann et al[19] identified six distinct molecular subtypes of TNBC: Two basal-like (BL1 and BL2), immunomodulatory (IM), mesenchymal (M), mesenchymal stem-like and luminal androgen receptor (LAR). Among these TNBC molecular subtypes, BL1, BL2, IM and M are predominantly basal-like[20]. This sub-classification of TNBC is clinically relevant due to the differential clinical outcome and chemotherapeutic response of each TNBC subtype[20,21]. Two lines of evidence indicate that distinct molecular subtypes of TNBC manifest differential responses to immunoediting, the process by which tumor cells escape the anti-tumor effect of immunosurveillance[22,23]. Similar to Lehmann’s sub-classification of TNBC, Burstein et al[22] utilized RNA and DNA profiling analysis to define TNBC subtypes as: (1) LAR; (2) mesenchymal (MES); (3) basal-like immunosuppressed (BLIS); and (4) basal-like immune-activated (BLIA). BLIS tumors have the worst prognosis and BLIA tumors manifest the best[22], in part due to the ability of the immune system to target them. Similarly, Jézéquel et al[23] sub-classified TNBC into three different subtypes via gene-expression profiles, including the C1 subtype (LAR, 22%), the C2 subtype (basal-like with a low immune response and high M2-like macrophages, 45%) and the C3 subtype (basal-enriched with a high immune response and low M2-like macrophages, 33%).

zéquel et al[23] sub-classified TNBC into three different subtypes via gene-expression profiles, including the C1 subtype (LAR, 22%), the C2 subtype (basal-like with a low immune response and high M2-like macrophages, 45%) and the C3 subtype (basal-enriched with a high immune response and low M2-like macrophages, 33%). They also found that basal-enriched C3 with a high immune response had a better prognosis than basal-like C2 with a low immune response[23]. While the molecular mechanisms leading to immune tolerance are not fully understood, two lines of study indicate that inactivation of tumor suppressor p53 and activation of CUL4A E3 ubiquitin ligase are involved in failure of tumor immunosurveillance[24,25].

prognosis than basal-like C2 with a low immune response[23]. While the molecular mechanisms leading to immune tolerance are not fully understood, two lines of study indicate that inactivation of tumor suppressor p53 and activation of CUL4A E3 ubiquitin ligase are involved in failure of tumor immunosurveillance[24,25]. TN/BLBC are more sensitive to preoperative or neoadjuvant chemotherapy than luminal breast cancers. Current therapeutic options for TN/BLBC include cytotoxic (e.g., combined treatment with anthracyclines or taxanes) and targeted (e.g., PARP1 and EGFR inhibition) therapies[20,21,26]. Although cytotoxic therapeutics achieves good tumor regression rates in the neo-adjuvant setting, patients experience frequent recurrence in five years after treatment[26,27]. Targeted therapies have also encountered discrepancies in trial results and issues with resistance[26,27]. TN/BL drug resistance potentially involves several mechanisms including intrinsic therapy resistance by a minor cell population in tumors, therapy-induced senescence and polyploidy in tumor cells, acquired therapeutic resistance via upregulation of drug efflux transporters, and acquired resistance via genetic reversion[28]. It is well accepted that tumor-initiating cells [generally called cancer stem cells (CSCs), discussed in the following paragraph] have high intrinsic drug-resistance, and may lead to relapse[28]. In regard to cell-cycle-related mechanisms, Puig et al[29] reported that cisplatin treatment induced senescent giant polyploid cells via DNA endoreduplication. These giant, multi-nucleated cells were able to generate small-sized, diploid cells that started to proliferate and were increasingly cisplatin-resistant[29]. These findings suggest that the multistep drug-resistant progression in which cells undergo DNA endoreduplication, polyploidization, depolyploidization and then generation of clonogenic escape cells, can account for tumor relapse after initial efficient chemotherapy[29]. Another common mechanism of drug resistance in cancer is the upregulation of ATP-binding cassette (ABC) transporter family proteins, which increases the efflux of chemotherapeutic drugs[28]. Besides these mechanisms, genetic alterations to restore the function of DNA repair proteins (e.g., BRCA1, BRCA2, FANCA, etc.) have been identified as a novel drug-resistant mechanism in DNA-repair-deficient cancers, which are initially sensitive to DNA-damaging agents (e.g., cisplatin) and to PARP inhibitors[28].

28]. Besides these mechanisms, genetic alterations to restore the function of DNA repair proteins (e.g., BRCA1, BRCA2, FANCA, etc.) have been identified as a novel drug-resistant mechanism in DNA-repair-deficient cancers, which are initially sensitive to DNA-damaging agents (e.g., cisplatin) and to PARP inhibitors[28]. In addition to these general resistance mechanisms, dysregulation of signal pathway regulators in basal-like breast cancers have recently been identified to be responsible for resistance to neoadjuvant chemotherapy and PARP inhibitor treatment. For example, basal-like breast cancers have low expression of dual specificity protein phosphatase 4 (DUSP4), which negatively regulates the Ras-ERK pathway, due to hypermethylation of the DUSP4 promoter. This results in activation of the Ras-ERK pathway and resistance to neoadjuvant chemotherapy[30]. Furthermore, overexpression of p38 mitogen-activated protein kinase in basal-like breast cancers enabled cancer cells to be resistant to the treatment of PARP inhibitors[31]. These results show the promise of increased molecular and functional characterization of TNBC subtypes. While the development of therapeutics to effectively treat TNBC still faces tremendous challenges, molecular targeting of TNBC based on molecular subtypes is a promising therapeutic strategy.

of PARP inhibitors[31]. These results show the promise of increased molecular and functional characterization of TNBC subtypes. While the development of therapeutics to effectively treat TNBC still faces tremendous challenges, molecular targeting of TNBC based on molecular subtypes is a promising therapeutic strategy. A conceptual model has been proposed that describes the developmental process of breast cancer as the progression of atypical hyperplasia into carcinoma in situ and finally to invasive carcinoma[32]. To better understand the relationship between DCIS and invasive breast carcinoma in breast cancer progression and to molecularly classify DCIS lesions, immunohistochemical and gene expression profiling analyses were used to define DCIS subtypes. The molecular subtypes of invasive breast cancer are also present at the DCIS stage, although their frequencies are varied between these two distinct breast cancer stages[33-35]. Furthermore, genetic studies of DCIS and invasive breast carcinoma have demonstrated that they have remarkable similarity in their genetic profiles when matched by histological grade and hormone receptor status[3,36-38]. These findings strongly support the theory that DCIS is a non-obligate precursor of invasive breast cancer. The distinct tumor subtypes may be generated from different cells of origin, by distinct tumor progression pathways, or a combination of both events. The ability of DCIS to progress to invasive disease is a complex biological phenomenon and depends on multiple factors, including genetic/epigenetic aberrations, genomic instability and the stromal microenvironment[3,32].

ifferent cells of origin, by distinct tumor progression pathways, or a combination of both events. The ability of DCIS to progress to invasive disease is a complex biological phenomenon and depends on multiple factors, including genetic/epigenetic aberrations, genomic instability and the stromal microenvironment[3,32]. As poorly differentiated invasive ductal carcinomas (IDC), basal-like breast tumors presumably have a DCIS precursor with similar cytologic and immunophenotypic features. Several studies reported that the use of both gene expression profiling and immunohistochemical analysis of basal-specific protein markers identified DCIS tumors with molecular features of basal-like invasive breast cancer[33,39-43]. These basal-like DCIS (BL-DCIS) tumors are presumed to be precursors for basal-like IDC (BL-IDC)[39,44]. Identification of BL-DCIS sheds light on the possibility of preventing progression to malignant basal-like breast cancer through the therapeutic targeting of precursor DCIS lesions. This review summarizes the recent investigation of BL-DCIS characteristics and the potential precursor relationship between BL-DCIS and invasive basal-like breast carcinoma.

ht on the possibility of preventing progression to malignant basal-like breast cancer through the therapeutic targeting of precursor DCIS lesions. This review summarizes the recent investigation of BL-DCIS characteristics and the potential precursor relationship between BL-DCIS and invasive basal-like breast carcinoma. CSCs have been identified in many types of cancer including breast cancer, and have significantly changed the strategy of cancer therapy. CSCs are a small tumor cell subpopulation that can be identified by several methods including fluorescence-activated cell sorting (FACS) analysis of stem/progenitor-cell-specific surface protein markers, aldehyde dehydrogenase (ALDH) activity assays, and FACS analysis of the “side population” indicated by Hoechst dye exclusion. CSCs have the unique ability to self-renew and to differentiate into heterogeneous tumor cell lineages in vitro and in vivo[45-48]. CSCs possess higher tumorigenic ability than non-CSCs, which can be measured by in vivo xenograft tumor formation assays involving the injection of enriched CSC fractions to immunodeficient mice[45-48]. Similar to normal stem cells, CSCs manifest stem-cell-specific gene expression signatures and can undergo symmetric as well as asymmetric cell division[45-48]. CSCs also exhibit high levels of drug resistance. Under the stress of chemotherapy they can become quiescent and resistant to drugs that target proliferating cells[49,50]. In addition, CSCs generally express high levels of multi-drug resistant ABC transporters and pump out anti-cancer drugs at high rates[51]. The stem-like characteristics of CSCs are due to dysregulation of stemness signaling pathways, such as the Notch, hedgehog, Wnt, TGF-β and pluripotent transcription factor (e.g., SOX2, OCT4, KLF4, etc.) pathways[52,53]. Moreover, CSCs possess similar characteristics to cells that have undergone epithelial-to-mesenchymal transition (EMT), allowing CSCs to survive in circulation and contribute to the metastasis of invasive cancers[54,55]. Due to these traits, CSCs are believed to be necessary for tumor heterogeneity, relapse, metastasis, and drug resistance. Breast CSCs were first identified in primary breast carcinomas using the markers CD44+/CD24−[56]. This minor tumor cell subset can self-renew to form tumorspheres in in vitro suspension culture conditions, and exhibit stem-cell gene expression patterns[56].

tumor heterogeneity, relapse, metastasis, and drug resistance. Breast CSCs were first identified in primary breast carcinomas using the markers CD44+/CD24−[56]. This minor tumor cell subset can self-renew to form tumorspheres in in vitro suspension culture conditions, and exhibit stem-cell gene expression patterns[56]. Isolated CD44+/CD24− cells have a higher capacity to initiate in vivo xenograft mammary tumors compared with other cell subsets and can differentiate into heterogeneous breast tumor cell lineages[56]. Furthermore, breast CSCs display a drug-resistant phenotype, and are able to better tolerate anti-cancer drug treatment than non-CSCs[57,58]. Enrichment of CSCs in breast cancer correlates with tumor aggressiveness, likely due to the characteristics described above. Among the molecularly-classified breast cancer subtypes, basal-like breast carcinomas tend to possess the highest proportion of CSCs compared with other subtypes, consistent with their heterogeneity and aggressiveness[59,60]. CSCs in basal-like breast cancer have emerged as a key target for cancer therapy.

ribed above. Among the molecularly-classified breast cancer subtypes, basal-like breast carcinomas tend to possess the highest proportion of CSCs compared with other subtypes, consistent with their heterogeneity and aggressiveness[59,60]. CSCs in basal-like breast cancer have emerged as a key target for cancer therapy. These discoveries have prompted cancer researchers to investigate the existence of CSCs and their characteristics in BL-DCIS[61,62]. Studies employing in vitro cell line and in vivo xenograft tumor models have significantly propelled the understanding of CSC characteristics and their role in BL-DCIS. They also give new insights into the aberrant molecular mechanisms involved in regulating CSC formation in BL-DCIS and the BL-DCIS-to-IDC transition. This article will review recent advances in these topics and their translational implications to the prognosis and prevention of BL-DCIS progression to invasive basal-like breast cancer.

ew insights into the aberrant molecular mechanisms involved in regulating CSC formation in BL-DCIS and the BL-DCIS-to-IDC transition. This article will review recent advances in these topics and their translational implications to the prognosis and prevention of BL-DCIS progression to invasive basal-like breast cancer. EXISTENCE AND FEATURES OF BASAL-LIKE-DCIS Due to the lack of effective targeted therapies, invasive basal-like breast cancers have poor prognosis. This has prompted cancer researchers to investigate the existence of precursor DCIS lesions that can potentially develop into BL-IDC. If precursor DCIS lesions with the potential to develop into BL-IDC are identified, patients with these lesions can be treated earlier with more aggressive therapies to prevent tumor progression and recurrence. Precursor DCIS lesions are presumed to have cytologic and immunophenotypic features similar to BL-IDC. By characterizing protein markers such as ER, PR, HER2, basal cytokeratins (e.g., CK5/6, CK14, and CK17), EGFR, c-kit and p63, about 6%-8% of DCIS cases were identified to be TN/BLBC[39-41,43]. In addition to immunohistochemical surrogates, Hannemann et al[33] performed microarray-based gene expression profiling to analyze and classify 40 in situ and 40 invasive breast cancer cases. Their two-dimensional hierarchical clustering analysis of microarray data showed that the luminal, HER2 and basal-like subtypes originally described in invasive breast cancer could also be identified in DCIS. A population-based cohort has shown that patients with BL-DCIS have a higher risk for local recurrence and development into invasive cancer compared with other molecular subtypes[63], demonstrating the need for further molecular characterization. The BL-DCIS subtype is associated with unfavorable prognostic variables such as high-grade nuclei, mutant p53 overexpression and elevated Ki-67 index[42]. In addition, through RNA deep sequencing analysis, Abba et al[64] subdivided high-grade DCIS into two subtypes, DCIS-C1 and DCIS-C2. The more aggressive DCIS-C1 (highly proliferative, basal-like, or ERBB2+) had a molecular signature characteristic of activated regulatory T (Treg) cells (CD4+/CD25+/FOXP3+) and CTLA4+/CD86+ complexes, indicative of a tumor-associated immunosuppressive phenotype[64]. This is the first evidence identifying mechanisms of immune evasion in BL-DCIS. Recently BL-DCIS tumors have also been associated with cell cycle-related biomarkers[65,66].

tivated regulatory T (Treg) cells (CD4+/CD25+/FOXP3+) and CTLA4+/CD86+ complexes, indicative of a tumor-associated immunosuppressive phenotype[64]. This is the first evidence identifying mechanisms of immune evasion in BL-DCIS. Recently BL-DCIS tumors have also been associated with cell cycle-related biomarkers[65,66]. Over 80% of BL-DCIS cases were p16-positive, whereas over 90% of DCIS cases with the luminal A phenotype were p16-negative[66]. In addition to p16 expression, co-expression signatures of p16+/Ki67+/COX2+ and p16+/Ki67+/COX2− were also found to be associated with the basal phenotype in DCIS and IDC[66]. These cell-cycle related profiles could be exploited to guide more aggressive treatment strategies in patients with high-grade DCIS. According to studies by Tamimi et al[42], the frequency (7.7%) of BL-DCIS in diagnosed DCIS cases is slightly lower than that (10.7%) of BL-IDC in diagnosed invasive breast cancer cases. One plausible explanation for the slightly higher frequency of basal-like expression in high-grade invasive vs in situ tumors is either that BL-DCIS lesions rapidly progress, leading to the lower identifiable frequency, or that the basal-like phenotype is acquired during invasive progression. Studies investigating the precursor potential of comedo-DCIS tumors (comedo-DCIS), a type of high-risk in situ breast lesions, identified a novel p63/CK5/Her2/neu-expressing cell subpopulation with ER-/PgR-/EGFR-[67]. Given that p63 alone and p63/Her2/neu co-expression are both associated with microinvasion and the recurrence of clinical comedo-DCIS, the p63/Her2/neu-expressing precursor intermediate is considered a cellular basis for the emergence of p63+/Her2/neu− or p63+/Her2/neu+ basal-like breast cancer, and thus may serve as a biomarker for identifying the BL-DCIS subgroup[67].

are both associated with microinvasion and the recurrence of clinical comedo-DCIS, the p63/Her2/neu-expressing precursor intermediate is considered a cellular basis for the emergence of p63+/Her2/neu− or p63+/Her2/neu+ basal-like breast cancer, and thus may serve as a biomarker for identifying the BL-DCIS subgroup[67]. CSCS AND BASAL-LIKE-DCIS Evidence of CSCs existing in DCIS It has been proposed that CSCs are responsible for generating tumor heterogeneity and the malignant progression of cancer. To validate this hypothesis within breast cancer and investigate the mechanisms of the DCIS to IDC transition, researchers are seeking to identify the existence of CSCs in DCIS and characterize their cell properties. To study the heterogeneous tumorgenicity of cancer cells, Damonte et al[68] generated mammary intraepithelial neoplasia (MIN) outgrowth lines. Derived from premalignant atypical lesions from PyV-mT transgenic mice, the MIN outgrowth lines are able to grow orthotopically in cleared mammary fat pads, and form a mammary tumor structure similar to human DCIS.

ity of cancer cells, Damonte et al[68] generated mammary intraepithelial neoplasia (MIN) outgrowth lines. Derived from premalignant atypical lesions from PyV-mT transgenic mice, the MIN outgrowth lines are able to grow orthotopically in cleared mammary fat pads, and form a mammary tumor structure similar to human DCIS. The 6 MIN lines generated demonstrated a varied ability to progress to invasive carcinoma with pulmonary metastatic potential via serial transplantation[68], establishing the paradigm that pre-CSCs in DCIS are capable of self-renewal, multilineage differentiation, and serve as the origin of invasive cancer. Notably, their studies indicate that sequential genetic hits for malignant transformation are not required for this DCIS model to progress to invasive and metastatic mammary carcinoma, and the programmed potential for latency and metastasis might be predetermined in these pre-CSCs[68]. Moreover, Espina et al[69] studied ex vivo organoid culture of fresh human DCIS lesions without enzymatic digestion or sorting, and found that DCIS contains malignant precursor cells that were able to form spheroids and a duct-like 3D structure in ex vivo organoid culture and to exhibit tumorigenicity in NOD/SCID mice[69].

over, Espina et al[69] studied ex vivo organoid culture of fresh human DCIS lesions without enzymatic digestion or sorting, and found that DCIS contains malignant precursor cells that were able to form spheroids and a duct-like 3D structure in ex vivo organoid culture and to exhibit tumorigenicity in NOD/SCID mice[69]. Evidence for the presence of CSCs in BL-DCIS The two lines of evidence mentioned above raised the possible existence of tumorigenic CSCs in BL-DCIS, which may determine the phenotypes of BL-DCIS and the capability of BL-DCIS to progress into invasive cancer. To confirm this, our research group characterized the BL-DCIS cell model MCF10DCIS.COM, which is derived from the non-cancerous breast epithelial cell line MCF10A. This BL-DCIS-mimic cell model has a unique bipotent progenitor ability, and is able to generate both myoepithelial and luminal-type cells in vivo, giving rise to BL-DCIS with high similarities to human DCIS lesions[70-76]. In addition to the formation of DCIS-like tumor structures in vivo, these tumor lesions are able to spontaneously progress to invasive breast cancer[74,76]. In our studies of MCF10DCIS.COM, we identified a CSC population with enriched ALDH1+ and the molecular signature CD44+/CD49f+/CD24−[62]. Compared with the non-stem-like cell subset, these stem-like cells possessed enhanced migration, invasion and self-renewal capacity, and accelerated xenograft tumor growth in nude mice[62]. Pandey et al[61] have also identified CSCs in the MCF10DCIS.COM cell line using similar cell surface markers (CD44+/ESA+/CD24−). In line with our result, CSCs isolated using this profile showed significantly higher DCIS tumor-initiating ability compared with non-stem-like cells[61]. The existence of CSCs in BL-DCIS raises the possibility that this CSC population serves as a malignant precursor necessary for the progression of BL-DCIS to BL-IDC.

In line with our result, CSCs isolated using this profile showed significantly higher DCIS tumor-initiating ability compared with non-stem-like cells[61]. The existence of CSCs in BL-DCIS raises the possibility that this CSC population serves as a malignant precursor necessary for the progression of BL-DCIS to BL-IDC. DEREGULATED FACTORS INVOLVED IN THE GENERATION OF CSCS AND THE BL-DCIS-TO-BL-IDC TRANSITION A challenging question in the breast cancer research field is how precursor DCIS lesions progress to invasive breast carcinomas. Attempts to address this critical question are hampered by the complexity of heterogeneous DCIS lesions. To overcome this barrier, the aforementioned MCF10DCIS.COM cell line has been extensively exploited as a unique model to study DCIS and explore molecular mechanisms involved in regulating the progression of DCIS to IDC. As MCF10DCIS.COM belongs to the BL-DCIS subtype, we review recent findings of deregulated factors implicated in the generation of basal CSCs in this cell model, and in enhancing malignancies as well as invasive progression of this BL-DCIS model in vivo. Further characterization of dysregulated signaling pathways involved in CSCs would advance insights into how BL-DCIS tumors progress to invasive cancer and propel the development of effective therapeutics to prevent this malignant progression.

malignancies as well as invasive progression of this BL-DCIS model in vivo. Further characterization of dysregulated signaling pathways involved in CSCs would advance insights into how BL-DCIS tumors progress to invasive cancer and propel the development of effective therapeutics to prevent this malignant progression. The role of miR-140 MicroRNAs (miRNAs) are short non-coding RNA molecules with a length of approximately 22 nucleotides that bind to the 3′-untranslated region of messenger RNAs and regulate mRNA stability and/or translation. miRNAs have been extensively investigated in cancer and other diseases, and regulate a variety of physiological and pathological processes at the posttranscriptional level. Although numerous miRNAs have been found to be involved in regulating CSCs in breast cancer[77], the miRNAs participating in basal CSC regulation and the tumorigenic development of BL-DCIS remain largely unknown. Through miRNA profiling of paired DCIS tumors, we identified downregulation of miR-140 as a hallmark of BL-DCIS lesions[78]. Our studies have shown that miR-140 is a tumor-suppressive miRNA which targets the stem-cell related factor SOX9 for degradation in in normal breast epithelial cells[78]. The degree of miR-140 downregulation positively correlates with the increased expression of SOX9 and the grade of DCIS lesions, implicating the critical role of the miR-140/SOX9 axis in the progression of DCIS[78]. Our studies also revealed that miR-140 was downregulated in cancer stem-like CD44+/CD24− cells isolated from MCF10DCIS.COM cells compared with normal breast stem cells isolated from MCF10A cells[78]. Moreover, restoration of miR-140 expression in MCF10DCIS.COM cells suppressed CSC self-renewal, invasion and in vivo tumorigenicity[78]. This suggests that the miR-140/SOX9 regulatory circuit is pivotal for the self-renewal and invasive capacity of basal CSC and their tumor formation in vivo, and is a potential therapeutic target.

storation of miR-140 expression in MCF10DCIS.COM cells suppressed CSC self-renewal, invasion and in vivo tumorigenicity[78]. This suggests that the miR-140/SOX9 regulatory circuit is pivotal for the self-renewal and invasive capacity of basal CSC and their tumor formation in vivo, and is a potential therapeutic target. The role of the nuclear receptor coactivator amplified in breast cancer 1 Ory et al[79] found that expression of nuclear receptor coactivator amplified in breast cancer 1 (AIB1) was aberrantly upregulated in DCIS lesions compared with normal breast. AIB1 activates NOTCH, HER2 and HER3 signaling pathways in MCF10DCIS.COM cells, and is required for the malignant phenotype of MCF10DCIS.COM in 3D culture and in vivo tumor formation and progression[79]. Critically, AIB1 inhibition led to a significant reduction in the CD44+/CD24− CSC population and also resulted in decreased myoepithelial progenitor cells in DCIS lesions in vitro and in vivo[79]. These data indicate that activation of AIB1 is an aberrant mechanism that initiates and maintains DCIS in vivo by facilitating the development as well as maintenance of basal CSCs. It is likely that aberrantly activated AIB1 assists other deregulated factors to promote the transition of BL-DCIS to BL-IDC.

in vivo[79]. These data indicate that activation of AIB1 is an aberrant mechanism that initiates and maintains DCIS in vivo by facilitating the development as well as maintenance of basal CSCs. It is likely that aberrantly activated AIB1 assists other deregulated factors to promote the transition of BL-DCIS to BL-IDC. The role of the p63-membrane-type 1-matrix metalloproteinase axis A critical step in the progression from DCIS to the invasive lesion is the crossing of the basement membrane and invasion into the stroma. This is achieved through degradation of extracellular matrix (ECM) proteins in the basement membrane by membrane-anchored matrix metalloproteinases (MMPs), including membrane-type 1 (MT1)-MMP[80]. MT1-MMP has been shown to be involved in invasive tumor growth and metastasis in several experimental cancer models[81-84]. Lodillinsky et al[85] analyzed expression of MT1-MMP in a large cohort of DCIS, IDC and microinvasive breast tumors, and found that MT1-MMP was significantly upregulated in the DCIS to IDC transition, and correlated with higher grade and hormone receptor-negative tumors. Functional analysis showed that silencing of MT1-MMP in MCF10DCIS.COM cells impaired the ability of this DCIS tumor model to progress into infiltrating lesions in vivo[85]. Additionally, Lodillinsky et al[85] identified p63 as an upstream positive regulator that increases MT1-MMP expression in DCIS, and is required for activating the basement membrane-invasive program of DCIS. Their findings suggest that aberrant activation of the p63/MT1-MMP axis in DCIS may contribute to the progression of DCIS to high-grade basal-like breast cancers. Although their studies did not address the role of the p63/MT1-MMP axis in MCF10DCIS.COM CSCs, p63 is a well-known basal-associated molecular marker, and has been recently found to be elevated in CSCs of HER2-type breast cancer and essential for their self-renewal as well as tumorigenicity[86]. Therefore, their results imply that aberrant activation of the p63/MT1-MMP axis in basal CSCs is a potential mechanism to trigger the progression of BL-DCIS to BL-IDC.

arker, and has been recently found to be elevated in CSCs of HER2-type breast cancer and essential for their self-renewal as well as tumorigenicity[86]. Therefore, their results imply that aberrant activation of the p63/MT1-MMP axis in basal CSCs is a potential mechanism to trigger the progression of BL-DCIS to BL-IDC. The role of Singleminded-2s Recent work has demonstrated that the basic helix-loop-helix/PER-ARNT-SIM (bHLH/PAS) transcription factor Singleminded-2s (SIM2s) is critical for normal mammary gland development and promoting tumor cell differentiation[87]. SIM2s is inhibited by C/EBPβ and NOTCH, important promoters of EMT and cell differentiation. Loss of SIM2s enhances EMT in the mouse mammary gland, normal breast and breast cancer cell lines. Moreover, SIM2s is frequently downregulated in human breast cancer. When SIM2s expression was restored in human breast cancer cell lines, their proliferation and invasion were suppressed. These results suggest that SIM2s is a tumor suppressor gene that is crucial for maintaining epithelial integrity through inhibiting the EMT program and promoting cell differentiation. To address the role of SIM2s in the transition of DCIS-to-IDC, Scribner et al[87] analyzed SIM2s expression in MCF10DCIS.COM and found that it is downregulated in this DCIS cell model when compared with non-cancerous MCF10A cells. Moreover, their functional studies showed that reestablishment of SIM2s in MCF10DCIS.COM cells significantly impaired their growth and invasion both in vitro and in vivo by promoting tumor cell differentiation. This is characterized by increased expression of luminal markers, including β-casein, E-cadherin, keratin 18, and decreased expression of genes associated with stem cell maintenance and a basal/EMT phenotype, including smoothened, p63, Snail-2, keratin 14 and vimentin[87]. In contrast, abrogation of SIM2s in MCF10DCIS.COM-derived xenograft tumors led to a more invasive phenotype and increased lung metastasis, correlating with the elevated expression of Hedgehog signaling and MMP[87]. From our and other studies indicating that basal CSCs are the origin of the tumorigenic and invasive characteristics of MCF10DCIS.COM cells[61,62,79], it is likely that decreased expression of SIM2s promotes the development of basal CSCs in BL-DCIS and further reduction in its expression activates invasive features of CSCs to facilitate the invasive progression of BL-DCIS into invasive breast carcinoma.

c and invasive characteristics of MCF10DCIS.COM cells[61,62,79], it is likely that decreased expression of SIM2s promotes the development of basal CSCs in BL-DCIS and further reduction in its expression activates invasive features of CSCs to facilitate the invasive progression of BL-DCIS into invasive breast carcinoma. The role of lipogenesis Cancer cells have altered metabolisms in comparison to normal cells[88], and upregulation of lipogenic genes and increased lipogenesis are hallmarks of late-stage breast cancer[89]. Inhibition of key lipogenic enzymes results in suppression of tumorigenicity both in vitro and in vivo by blocking proliferation and inducing apoptosis[90-93]. Although the role of increased lipogenesis in late-stage breast cancer has been extensively studied, its role in early-stage breast cancer DCIS still remains elusive. Moreover, whether lipogenesis is engaged in regulating CSCs of DCIS is an interesting yet unexplored question. To address the role of lipogenesis in DCIS CSCs, Pandey et al[61] used the cell surface marker profile (CD44+/ESA+/CD24−) to isolate CSCs from the MCF10DCIS.COM cell line for expression analysis of lipogenic genes. Their studies showed that expression levels of all lipogenic genes tested in the CSC population were significantly higher than the normal stem-like counterpart population isolated from non-cancerous MCF10A cells. To further investigate the role of sterol regulatory element-binding protein-1 (SREBP1), the master transcriptional activator of lipogenic genes, in CSCs, SREBP1 was ectopically overexpressed in MCF10A stem-like cells. Overexpression of SREBP1 caused enhanced lipogenesis, cell growth and mammosphere formation[61]. When upregulated in MCF10AT, a MCF10A-derived, premalignant cell line, SREBP1 promoted DCIS generation in vivo by increasing CSC survival[61]. These findings indicate that activation of lipogenesis is a pre-requisite for basal CSC generation and DCIS formation, and is important for endowing increased cell survival capacity.

upregulated in MCF10AT, a MCF10A-derived, premalignant cell line, SREBP1 promoted DCIS generation in vivo by increasing CSC survival[61]. These findings indicate that activation of lipogenesis is a pre-requisite for basal CSC generation and DCIS formation, and is important for endowing increased cell survival capacity. THE IMPACTS OF THE TISSUE MICROENVIRONMENT ON CSCS OF BL-DCIS DCIS lesions are heterogeneous tumors encapsulated by the myoepithelium and basement membrane. When they progress to IDC, tumor cells cross the myoepithelial layer and basement membrane and invade into the stroma, comprised of stromal fibroblasts/preadipocytes, mature adipocytes, immune and endothelial cells. Therefore, these various tissue cells and the ECM that composes the tumor microenvironment can regulate CSC self-renewal and differentiation, DCIS formation, progression into invasive lesions, and metastasis[94-98]. Understanding the impact of the tumor microenvironment on basal CSCs and BL-DCIS could potentially enable the design of more effective diagnosis and intervention strategies to improve the survival of cancer patients. There are two lines of recent studies indicating the critical impacts of the tissue microenvironment on the tumorigenesis of BL-DCIS and their transition to BL-IDC, exosomal signaling and ECM dependent signaling.

design of more effective diagnosis and intervention strategies to improve the survival of cancer patients. There are two lines of recent studies indicating the critical impacts of the tissue microenvironment on the tumorigenesis of BL-DCIS and their transition to BL-IDC, exosomal signaling and ECM dependent signaling. Exosomal signaling from the tumor microenvironment Exosomal secretion is a newly identified mechanism of paracrine signaling through which cells secret exosomes, microvesicles with a diameter usually less than 100 nm, which can contain cargo proteins, nucleic acids and nutrients[99]. Secreted exosomes can transduce their carried contents into surrounding cells via the cell internalization mechanism mediated by the heparan sulfate proteoglycan receptors[99]. The known roles of exosomes in tumorigenesis are to restructure the tumor tissue microenvironment, modulate tumor immune responses and directly regulate tumor cell behaviors via their delivery of proteins and genetic materials. miRNAs have been found to be one kind of nucleic acids carried by secreted exosomes[99,100]. Given that miRNAs are regulatory factors that can modulate protein expression, exosomal trafficking of miRNAs has been recognized to be a microenvironmental signal that can affect signaling networks at the post-transcriptional level[100]. From BL-DCIS studies, we found that the miRNA content in exosomes secreted from CSCs of DCIS was altered compared to exosomes from normal stem-like breast cells[62]. Notably, CSC-secreted exosomes carried less miR-140, an aforementioned tumor-suppressive miRNA, than those secreted from normal stem-like cells, suggesting that the tumorigenic process alters exosomal contents[62].

somes secreted from CSCs of DCIS was altered compared to exosomes from normal stem-like breast cells[62]. Notably, CSC-secreted exosomes carried less miR-140, an aforementioned tumor-suppressive miRNA, than those secreted from normal stem-like cells, suggesting that the tumorigenic process alters exosomal contents[62]. In addition to the role of exosomal trafficking in signaling among DCIS tumor cells, we recently found that exosomes secreted from preadipocytes, the precursors of mature adipocytes, could impact the stemness and tumorigenic properties of CSCs in BL-DCIS[101]. Preadipocyte-derived exosomes enhanced in vitro cell migration as well as self-renewal of BL-DCIS cells and facilitated the xenograft tumor formation of transplanted BL-DCIS cells in vivo[101]. The enhanced effect of preadipocyte-secreted exosomes on the tumorigenicity of BL-DCIS might be attributable to a number of growth-promoting cytokines identified within these exosomes[101]. Taken together, these findings demonstrate that exosomal signaling plays an important role in the tumor microenvironment.

vo[101]. The enhanced effect of preadipocyte-secreted exosomes on the tumorigenicity of BL-DCIS might be attributable to a number of growth-promoting cytokines identified within these exosomes[101]. Taken together, these findings demonstrate that exosomal signaling plays an important role in the tumor microenvironment. ECM-dependent regulatory signaling High tumor heterogeneity correlates with poor prognosis due to its association with malignancies, recurrence, metastasis and anti-cancer drug resistance[102]. Intratumor heterogeneity could result from an intrinsic stochasticity in gene expression and from genetic and/or heritable epigenetic differences among tumor cells[103]. By studying the effect of ECM on an immortalized basal-like breast epithelial cell line, Wang et al[104] identified the ECM-dependent TGFBR3 (transforming growth factor β receptor 3)-JUND (jun D proto-oncogene)-KRT5 (keratin 5) regulatory circuit that generates heterogeneous gene expression among ECM-attached breast cells. This circuit is composed of two anticorrelated gene expression programs that negatively regulate each other. TGFBR3 signaling downregulates JUND mRNA levels, whereas JUND represses both TGFBR3 and JUND mRNA levels[104]. Perturbing this regulatory circuit in breast epithelial cells could lead to the formation of aberrant tissue lesions similar to high-grade DCIS[104]. Their studies also indicate that the TGFBR3-JUND circuit is the molecular mechanism responsible for the heterogeneous expression of KRT5 in some basal-like premalignant lesions[104]. These findings suggest that heterogeneous KRT5 expression patterns present in high-grade basal-like DCIS lesions are likely due to loss of tissue-level regulation of gene oscillatory networks rather than genetic selection. Disrupting the dependence of this regulatory circuit on ECM results in detachment of breast epithelial cells from the ECM, in turn leading to cell death[104]. However, some cells survive through activation of a juxtacrine tenascin C (TNC) deposition mechanism. TNC is a critical survival factor for detached cells that would otherwise be subjected to keratinization-induced or anoikis-dependent cell death, and participates in stabilizing the heterogeneous JUND-KRT5 expression[102,104]. These results are in line with the previous finding that metastasizing breast cancer cells express TNC to elicit and/or maintain their metastasis-initiating characteristics[105].

d to keratinization-induced or anoikis-dependent cell death, and participates in stabilizing the heterogeneous JUND-KRT5 expression[102,104]. These results are in line with the previous finding that metastasizing breast cancer cells express TNC to elicit and/or maintain their metastasis-initiating characteristics[105]. Particularly, it has been shown that TNC is able to increase the expression of stem-cell signaling proteins, suggesting its role in modulating the CSC population[105]. Their findings demonstrate that this ECM-dependent regulatory circuit program can maintain normal tissue architecture and function in addition to preventing cell outgrowth and migration when it is properly regulated. However, when dysregulated (e.g., aberrant ECM signaling), this system enables cells to evade keratinization and anoikis, and allows them to metastasize.

regulatory circuit program can maintain normal tissue architecture and function in addition to preventing cell outgrowth and migration when it is properly regulated. However, when dysregulated (e.g., aberrant ECM signaling), this system enables cells to evade keratinization and anoikis, and allows them to metastasize. THE IMPLICATIONS OF CSCS IN PROGNOSIS AND PREVENTION OF EARLY STAGE BASAL-LIKE BREAST CANCER The identification of CSCs in BL-DCIS opens a window for cancer researchers to explore how BL-DCIS initiate and progress into invasive basal-like breast cancer. In addition to promoting our understanding of the role of basal CSCs in BL-DCIS, these research advances have tremendous translational implications for future prognostic and therapeutic applications. The dysregulated molecular factors which result in basal CSC generation could potentially be exploited as prognostic biomarkers for BL-DCIS. This would help identify and grade the probability of diagnosed DCIS developing into BL-IDC, and determine whether DCIS patients should be treated more aggressively. If this prognostic system can be established, it will substantially benefit patients with DCIS and lower their chances of basal-like invasive breast cancer recurrence. The therapeutic agents that can target these deregulated factors could be potentially exploited for targeted therapy of DCIS. This chemopreventive strategy would save breast cancer patients’ lives, especially since there are currently no effective therapies to cure basal-like invasive breast cancer. Promising therapeutic agents have already been identified that are effective in targeting CSCs in BL-DCIS. The dietary compound sulforaphane (SFN) can restore miR-140 expression and downregulate the expression of miR-140 targets SOX9 and ALDH1, inhibiting the self-renewal of basal CSCs and DCIS formation in vivo[62]. Besides SFN, our studies of the chemopreventive agent Shikonin (SK), a bioactive compound found in the herbal plant shikon, showed that exosomes secreted from SK-treated preadipocytes lost the ability to promote BL-DCIS tumorigenicity both in vitro and in vivo. This is a novel chemopreventive mechanism for BL-DCIS, targeting the tumor microenvironment in place of the tumor itself[101]. Furthermore, a study from Watabe’s research group shows that resveratrol, a therapeutic agent capable of blocking the lipogenic gene expression in basal CSCs, is able to significantly suppress DCIS formation in animals[61].

e mechanism for BL-DCIS, targeting the tumor microenvironment in place of the tumor itself[101]. Furthermore, a study from Watabe’s research group shows that resveratrol, a therapeutic agent capable of blocking the lipogenic gene expression in basal CSCs, is able to significantly suppress DCIS formation in animals[61]. These exciting findings provide a strong rationale to propel the development of chemopreventive therapeutics for DCIS patients after surgical and radiological treatment.

e mechanism for BL-DCIS, targeting the tumor microenvironment in place of the tumor itself[101]. Furthermore, a study from Watabe’s research group shows that resveratrol, a therapeutic agent capable of blocking the lipogenic gene expression in basal CSCs, is able to significantly suppress DCIS formation in animals[61]. These exciting findings provide a strong rationale to propel the development of chemopreventive therapeutics for DCIS patients after surgical and radiological treatment. CONCLUSION Although the research efforts to combat invasive basal-like breast cancer have provided tremendous insights into this breast cancer subtype, we still have not identified effective therapeutic agents and strategies to cure this disease. Therefore, identifying and targeting the precursor of aggressive breast cancer is a promising direction to prevent the occurrence of this disease. BL-DCIS is an early stage breast cancer with a high risk of recurrence, and targeting it may prevent cancer recurrence and progression to invasive disease. As summarized and discussed in this review, numerous signaling pathways and factors have been identified as dysregulated in basal CSCs of BL-DCIS. Moreover, several chemopreventive agents have been tested to target these deregulated mechanisms. These studies suggest targeting CSCs in BL-DCIS as a potential strategy to inhibit the tumorigenicity of BL-DCIS and prevent the progression of BL-DCIS into BL-IDC. However it is critical to test the proof-of-principle of these chemopreventive strategies in clinical trials. Developing reliable BL-DCIS biomarkers will allow clinicians to design effective targeted therapies that can prevent the recurrence and progression of early stage basal-like breast cancers.

BL-DCIS into BL-IDC. However it is critical to test the proof-of-principle of these chemopreventive strategies in clinical trials. Developing reliable BL-DCIS biomarkers will allow clinicians to design effective targeted therapies that can prevent the recurrence and progression of early stage basal-like breast cancers. Supported by The National Cancer Institute (NCI) of National Institutes of Health (NIH) of the United States of America to Zhou Q, Nos. 5R01CA157779-03 and 5R01CA163820-04. Author contributions: Lo PK wrote the manuscript; Wolfson B reviewed and edited the manuscript; Zhou Q designed the aim of the review and reviewed the manuscript. Conflict-of-interest statement: Authors declare no conflict of interests for this article. Core tip Basal-like ductal carcinoma in situ (BL-DCIS) often lacks endocrine receptors and has a high rate of recurrence due to no available targeted therapies. BL-DCIS is a precursor of invasive basal-like breast carcinoma, a malignant cancer prone to metastasis and drug-resistance. Therefore, targeting BL-DCIS to prevent transition into invasive cancer is of significant interest. The recent identification and characterization of cancer stem-like cells in BL-DCIS advance the understanding of BL-DCIS and their potential role in driving the progression of BL-DCIS to invasive basal-like breast cancer. These findings provide critical implications for the development of therapies that prevent the progression of BL-DCIS.

INTRODUCTION Women in the United States with advanced stage epithelial ovarian cancer (OC) have an overall 5-year survival rate of about 30%[1]. As with many cancers, survival is closely linked with the stage of diagnosis, such that women with localized (stage I) disease have a relative 5-year survival rate of 92%; the prognosis however declines with late stage disease and metastases[2]. Without an adequate early detection strategy, ensuring that women receive appropriate, standard of care (SOC) treatment is a very important intervention that has demonstrated reduction in OC mortality[3].

e a relative 5-year survival rate of 92%; the prognosis however declines with late stage disease and metastases[2]. Without an adequate early detection strategy, ensuring that women receive appropriate, standard of care (SOC) treatment is a very important intervention that has demonstrated reduction in OC mortality[3]. National Comprehensive Cancer Control Network (NCCN) current treatment recommendations for women with epithelial OC include an evaluation prior to initiating chemotherapy along with accurate surgical staging and primary debulking surgery/cytoreduction performed by a gynecologic oncologist (GO)[3]. In most but not all cases, at least six cycles of platinum and taxane-based chemotherapy administration is recommended for advanced epithelial OCs[3]. Appropriate care not only constitutes the receipt of SOC treatment, but also quality care from an experienced GO, who is trained to both perform the surgery and administer chemotherapy[3,4]. The evidence supporting better guideline-adherent care and outcomes among patients seen by a GO has been previously examined[5–8], and prior studies suggest only 30%–40% of women with OC are treated by a GO[5,9–11]. While NCCN cancer center patients tend to receive guideline-adherent care[12], there is potential in exploring whether differences in SOC treatment are affected across patient-level demographic and clinical subgroups.

ly examined[5–8], and prior studies suggest only 30%–40% of women with OC are treated by a GO[5,9–11]. While NCCN cancer center patients tend to receive guideline-adherent care[12], there is potential in exploring whether differences in SOC treatment are affected across patient-level demographic and clinical subgroups. To date, few studies have jointly considered surgical and chemotherapeutic SOC indicators in examining survival in OC patients[13–17]. In this study, we examine predictors of both SOC receipt (surgical and chemotherapeutic) and adherence to these treatments among women treated by GOs compared to non-GOs. We further quantified the survival advantage of SOC treatment receipt among OC patients. MATERIALS AND METHODS Data source and study population The study included all women in the Surveillance, Epidemiology, End Results (SEER)-Medicare database[18] diagnosed with OC from January 1, 1992 to December 31, 2006 (n = 38972). We excluded women who did not have a primary epithelial OC diagnosis (n = 6175); were Medicare age-ineligible (age < 66) at date of diagnosis (n = 11716); had an invalid month of diagnosis (n = 166); had diagnoses based on autopsy or death certificate only (n = 543); had a nonepithelial ovarian malignancy (n = 3198); and were not continuously enrolled in both Medicare Part A and B or were enrolled in an Health Maintenance Organization plan during the course of treatment (n = 5486). A total of 11688 OC patients met the inclusion criteria for the study.

th certificate only (n = 543); had a nonepithelial ovarian malignancy (n = 3198); and were not continuously enrolled in both Medicare Part A and B or were enrolled in an Health Maintenance Organization plan during the course of treatment (n = 5486). A total of 11688 OC patients met the inclusion criteria for the study. Definition of variables Patient-level covariates included age, race, stage at diagnosis, marital status, year of diagnosis, geographic region of SEER registry, and cancer histology. The Charlson-Klabunde comorbidity index score was determined using Medicare claims data for 12 mo prior to and 4 mo after cancer diagnosis date, per prior studies[19,20]. We examined all procedure codes in the Medicare claims data falling within a treatment window (defined as two months prior to and one year after the diagnosis date) to determine if a patient received surgical or chemotherapeutic SOC. Since only month and year of diagnosis are reported in the SEER database, the 15th day of the month was assigned as the day of diagnosis for each patient. SOC definitions Per recommendation from an experienced group of GOs, consulted specifically for this project (W. Brewster, R.E. Bristow and D.K. Singh), the International Federation of Gynecologists and Obstetricians (FIGO) stage of disease categories were grouped as: I A/I B, I C/II, IIIA/III B and IIIC/IV based on similarities in current surgical and chemotherapeutic treatment regimens. FIGO stage III NOS and stage IV were grouped into stage IIIC/IV group, given that a high proportion of all stage III cases were stage IIIC.

cians (FIGO) stage of disease categories were grouped as: I A/I B, I C/II, IIIA/III B and IIIC/IV based on similarities in current surgical and chemotherapeutic treatment regimens. FIGO stage III NOS and stage IV were grouped into stage IIIC/IV group, given that a high proportion of all stage III cases were stage IIIC. Among the women who met the inclusion criteria (n = 11688), we examined receipt of SOC among women receiving any initial surgical care. Thus, we further excluded women who received treatment outside of the treatment window (n = 28), those who had no procedure codes of interest for any surgical care (n = 2464), and women who received neoadjuvant chemotherapy (n = 2482) (given the difficulty of cancer staging for women who are eligible for neoadjuvant chemotherapy) to examine differences in guideline-adherent treatment and survival. We also excluded all OC patients diagnosed with stage I NOS or who were unstaged at diagnosis since minimum SOC parameters are not well defined for these groups.

the difficulty of cancer staging for women who are eligible for neoadjuvant chemotherapy) to examine differences in guideline-adherent treatment and survival. We also excluded all OC patients diagnosed with stage I NOS or who were unstaged at diagnosis since minimum SOC parameters are not well defined for these groups. The GO group defined minimum surgical SOC as lymph node dissection, omentectomy and oophorectomy for all patients with FIGO stage IA/IB, IC/II or IIIA/IIIB at diagnosis, but omentectomy and oophorectomy only for women with stage IIIC/IV at diagnosis. Minimum chemotherapy SOC definition depended on: (1) stage of disease at diagnosis; (2) number of chemotherapy cycles received; and (3) type of chemotherapy agent received. For analysis, chemotherapy SOC was defined as an individual receiving the defined number of cycles (three cycles of chemotherapy for stage IC/II and six cycles for stage III/IV), with at least one multi-agent cycle (defined as one platinum based and one non-platinum based agent) using either intravenous or intraperitoneal modes of administration. One cycle of chemotherapy was equal to three weeks of treatment, given that chemotherapy is usually administered every 3–4 wk[3,21]. Patients were documented as receiving overall SOC if they received both surgical and adjuvant chemotherapeutic SOC.

t) using either intravenous or intraperitoneal modes of administration. One cycle of chemotherapy was equal to three weeks of treatment, given that chemotherapy is usually administered every 3–4 wk[3,21]. Patients were documented as receiving overall SOC if they received both surgical and adjuvant chemotherapeutic SOC. The GO group recommended surgical and chemotherapy procedure codes for use in determining SOC for each FIGO stage category. Procedure codes included both International Classification of Diseases, Ninth revision, clinical modification codes and American Medical Association (AMA) Current Procedural Terminology codes. Surgeon specialty definition Self-reported, physician specialty information from the SEER-Medicare claims file was linked with and verified against the AMA Physician Masterfile using the unique provider identification number (UPIN) for physicians performing (or those in attendance) of an OC procedure of interest. If the operating physician UPIN was not available, but the attending physician UPIN was available, AMA specialty was assigned to the attending physician. If the UPIN for an operating and attending physician was unavailable, the self-reported physician specialty variable found in the Medicare data set was used to define specialty. When a patient received treatment from multiple physicians, care was attributed to the most specialized physician (most to least specialized: GO, gynecologist, general surgeon, and other physician). For analytic purposes, physician specialty was grouped as GO and non-GO.

Medicare data set was used to define specialty. When a patient received treatment from multiple physicians, care was attributed to the most specialized physician (most to least specialized: GO, gynecologist, general surgeon, and other physician). For analytic purposes, physician specialty was grouped as GO and non-GO. Statistical analysis We examined predictors associated with receipt of surgical and chemotherapeutic SOC. A forward selection logistic regression model was used to examine each question. Comparisons of the distribution of OC patients receiving the SOC by physician specialty was examined using the Pearson χ2 test.

Medicare data set was used to define specialty. When a patient received treatment from multiple physicians, care was attributed to the most specialized physician (most to least specialized: GO, gynecologist, general surgeon, and other physician). For analytic purposes, physician specialty was grouped as GO and non-GO. Statistical analysis We examined predictors associated with receipt of surgical and chemotherapeutic SOC. A forward selection logistic regression model was used to examine each question. Comparisons of the distribution of OC patients receiving the SOC by physician specialty was examined using the Pearson χ2 test. Cox proportional hazard methods were used to determine differences in survival time from date of OC diagnosis to date of death. The proportional hazards assumption was examined by testing interactions between time and each covariate in the model. The final models (Model 1 and 2) exclude women (n = 1003) who died within 4.5 mo after diagnosis (i.e., women who did not live long enough to receive chemotherapy SOC). Due to a common category in the chemotherapy variables (chemotherapy SOC and chemotherapy physician specialty), we examined two different models. The first model (Model 1) examined surgery physician specialty and receipt of both SOC measurements, while the second model (Model 2) examined both surgery and chemotherapy physician specialty and receipt of surgery SOC, adjusting for patient-level and clinical factors. All final models were adjusted for covariates that had a statistically significant association from the bivariate analysis or were of importance in the literature. All analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, NC, United States).

gery SOC, adjusting for patient-level and clinical factors. All final models were adjusted for covariates that had a statistically significant association from the bivariate analysis or were of importance in the literature. All analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, NC, United States). RESULTS Among the 11688 OC patients, 57.4% (n = 6714) received an initial surgical procedure code of interest. Table 1 shows the patient and tumor characteristics by the type of physician performing the initial surgery. The mean age of patients was mid to late-70s; most women were white, married or widowed, had no comorbidities, had FIGO stage IIIC/IV disease, and serous histology. More women received an initial surgical procedure from OB/GYNs (n = 3088) than GOs (n = 2254), general surgeons (n = 914), or other non-GO/unknown specialties (n = 419). Among women treated by a GO, 79.2% received the surgical SOC and 52.8% received the chemotherapy SOC (Figure 1). Regardless of stage at diagnosis, women more frequently received surgical and chemotherapeutic SOC from a GO than from a non-GO.

RESULTS Among the 11688 OC patients, 57.4% (n = 6714) received an initial surgical procedure code of interest. Table 1 shows the patient and tumor characteristics by the type of physician performing the initial surgery. The mean age of patients was mid to late-70s; most women were white, married or widowed, had no comorbidities, had FIGO stage IIIC/IV disease, and serous histology. More women received an initial surgical procedure from OB/GYNs (n = 3088) than GOs (n = 2254), general surgeons (n = 914), or other non-GO/unknown specialties (n = 419). Among women treated by a GO, 79.2% received the surgical SOC and 52.8% received the chemotherapy SOC (Figure 1). Regardless of stage at diagnosis, women more frequently received surgical and chemotherapeutic SOC from a GO than from a non-GO. Table 2 reports the factors associated with receipt of surgical SOC after adjusting for other covariates. Surgery performed by a GO was strongly associated with receiving surgical SOC [odds ratio (OR) for GO = 2.35; 95%CI: 2.03–2.71]. Other factors associated with greater odds of surgical SOC receipt included: More advanced stage of disease, white vs African-American race, younger age at diagnosis, serous vs adenocarcinoma not otherwise specified histologic type, being married vs not married, and diagnosis during the later years of the study period.

Other factors associated with greater odds of surgical SOC receipt included: More advanced stage of disease, white vs African-American race, younger age at diagnosis, serous vs adenocarcinoma not otherwise specified histologic type, being married vs not married, and diagnosis during the later years of the study period. Table 2 also reports factors associated with receipt of the minimum chemotherapy SOC after adjusting for other covariates. Women who obtained chemotherapy from a GO had a higher odds of receiving chemotherapeutic SOC (OR = 1.25, 95%CI: 1.07–1.47). Other statistically significant factors for higher odds of chemotherapeutic SOC included: Less advanced stage of disease, younger age at diagnosis, histologic type (serous compared with endometrioid/mucinous/clear cell), being married compared with unmarried, living in the SEER Midwest region (compared to the SEER Northeast), and diagnosis during more recent years.

her odds of chemotherapeutic SOC included: Less advanced stage of disease, younger age at diagnosis, histologic type (serous compared with endometrioid/mucinous/clear cell), being married compared with unmarried, living in the SEER Midwest region (compared to the SEER Northeast), and diagnosis during more recent years. Table 3 shows the Cox regression model of time to death among the sample of OC patients who received a primary surgery procedure who did not die within 4.5 mo after diagnosis. In Table 3 (Model 1), women who did not receive surgery SOC had increased mortality compared to women who did [hazard ratio 1.22 (95%CI: 1.12–1.33)]. Similarly, women who did not receive any chemotherapy SOC had a higher risk of earlier death compared to women who received the full contingent of chemotherapy [hazard ratio 1.29 (95%CI: 1.14–1.46)]. Increasing age, late stage disease, higher number of comorbidities, and mucinous histology compared to serous histology were all associated with increased death (Table 3, Model 1). Similar patterns were observed in Table 3, Model 2 after controlling for chemotherapy physician specialty (as opposed to chemotherapy SOC). For Model 2, women who received surgery from a GO had better survival. Although there was no significant difference in survival between chemotherapy treatment from a GO compared to non-GO, those not receiving any chemotherapy had a significantly shorter survival time (Table 3, Model 2). The median survival time for women who received the overall SOC was 52 mo compared to 38 mo for women that did not receive the overall surgical and chemotherapeutic SOC (Figure 2).

treatment from a GO compared to non-GO, those not receiving any chemotherapy had a significantly shorter survival time (Table 3, Model 2). The median survival time for women who received the overall SOC was 52 mo compared to 38 mo for women that did not receive the overall surgical and chemotherapeutic SOC (Figure 2). DISCUSSION Our findings show that among OC patients receiving initial surgical treatment, only 25% of women received the overall SOC as defined by our panel of GOs. Few women (approximately one-third of women receiving a surgical procedure) had a GO involved at any point during their care. Women who obtained surgery from a GO however, were more likely to receive the surgical SOC and chemotherapeutic SOC than women who obtained treatment from a non-GO. The median survival time was 14 mo longer for women who received the overall SOC compared to women who did not receive overall SOC.

oint during their care. Women who obtained surgery from a GO however, were more likely to receive the surgical SOC and chemotherapeutic SOC than women who obtained treatment from a non-GO. The median survival time was 14 mo longer for women who received the overall SOC compared to women who did not receive overall SOC. Our results are consistent with prior studies that suggest that appropriate surgical treatment in the United States is more frequently performed when a GO is the treating physician[5]. Data from a single state cancer registry study by Chan et al[14] showed that women with OC under the care of GOs were more likely to receive appropriate staging and chemotherapy treatments, controlling for age, stage, and grade of disease. Also similar to previous studies, our results suggest that greater utilization of GOs in the care of OC patients would be beneficial[22]. Although the level of detail in our analysis is unable to discriminate the factors underlying the low utilization, it is likely that our results reflect a complex interaction of both preference and access-relevant effects, such as the influence of a patient’s choice in receipt of GO care vs a shortage of available GOs in some areas.

evel of detail in our analysis is unable to discriminate the factors underlying the low utilization, it is likely that our results reflect a complex interaction of both preference and access-relevant effects, such as the influence of a patient’s choice in receipt of GO care vs a shortage of available GOs in some areas. While patient treatment preferences can independently and significantly affect chemotherapy receipt[23], geographic access may also play an important role in (both chemotherapeutic or surgical) treatment receipt from a GO. For example, a previous analysis reported on the unequal distribution of GOs in the United States[24]. A recently published study suggested that OC mortality may be a function of distance to a practicing GO as counties located more than 50 miles from a gynecologic oncology practice had almost 60% increased likelihood of OC mortality than those physically closer to a practice location[25]. While earlier research efforts have indicated that treatment of OC can be improved by early referral to a GO[5,10], referral and consultation from GOs have generally been low, with only about 39% of family physicians and 51% of general internists self-reporting referrals to a GO[26]. Given that surgery is an important determinant of outcomes for OC patients, receiving surgery/treatment from surgeons with specialized training in pelvic surgery (i.e., GOs)[27], who see a high volume of cases[10,28] at high volume facilities treating more than 20 OC cases per year[28,29], might help improve outcomes.

Given that surgery is an important determinant of outcomes for OC patients, receiving surgery/treatment from surgeons with specialized training in pelvic surgery (i.e., GOs)[27], who see a high volume of cases[10,28] at high volume facilities treating more than 20 OC cases per year[28,29], might help improve outcomes. It is important to note that there are still subgroups that require further research. Although African-American women were more likely than their white counterparts to receive their initial surgical procedure from a GO (data not shown), they had lower odds of receiving the surgical SOC and there was no difference in survival after adjusting for physician specialty, surgical SOC, and other tumor and sociodemographic characteristics. The increased risk of death among African American women noted in other studies, when controlling for receipt of chemotherapeutic SOC, suggests that there may be some important nuanced differences in the definitions of chemotherapeutic SOC[30,31], chemotherapeutic agents, and/or interaction effects between age, comorbidity, stage, and race that have not been adequately explored. Bristow et al[32] have previously suggested similar differences in survival between African-American and white OC patients and the complexity of examining race-based survival associations[33,34].

apeutic agents, and/or interaction effects between age, comorbidity, stage, and race that have not been adequately explored. Bristow et al[32] have previously suggested similar differences in survival between African-American and white OC patients and the complexity of examining race-based survival associations[33,34]. The findings in this study should be considered in light of several limitations: (1) our analysis was focused on fee for service Medicare; women who received treatment under managed care were not included because the managed care cases did not include codes to identify specific treatment procedures; (2) neoadjuvant chemotherapy cases, which could have later received surgical SOC, were excluded; and (3) it is a challenge to operationalize NCCN recommendations into an analytic/computer program because the recommendations are relatively complex, and some information required for the NCCN decision algorithms is not available in claims data. However, our panel of experienced GOs developed a simpler, but accurate definition of the SOC so that recommendations could be converted into analytic code. Similarly, since SOC definitions were varied for each stage at diagnosis, if claims data were not available for the full contingency of treatment procedures, it is possible that there was an underestimation of patients identified as receiving overall SOC in that subgroup. Fourth, given the limitations of Medicare data, inaccuracies or incomplete data in billing, drug, or procedure codes could have resulted in an underestimate or overestimate of the total number of surgeries and/or chemotherapy procedures performed, thus biasing the estimate. Previous studies have noted some concerns in the validation of chemotherapeutic agents within Medicare claims data[30,31]. Fifth, there is potential for misclassification of physician specialty, given the use of multiple data sources including operating physician, attending physician, and self-reported physician specialty[35]. Furthermore, in our analysis, receipt of treatment from a GO was designated as such if a GO had been seen at any point during the care. Lastly, since we assumed each cycle of treatment lasted three weeks, we calculated that it would take at least 4.5 mo for women diagnosed with stage IIIC or IV to complete the chemotherapy SOC as defined in our study. Thus, women who died within five months of the diagnosis date would not have had the opportunity to receive chemotherapy SOC.

d each cycle of treatment lasted three weeks, we calculated that it would take at least 4.5 mo for women diagnosed with stage IIIC or IV to complete the chemotherapy SOC as defined in our study. Thus, women who died within five months of the diagnosis date would not have had the opportunity to receive chemotherapy SOC. Our definition of chemotherapy SOC may have been too rigorous and potentially introduce selection or survival bias. Our study showed that GOs more often provided the surgical and chemotherapeutic SOC. The receipt of surgical standards was associated with better survival outcomes, even after adjusting for provider specialty. As such, these two NCCN-recommendations (i.e., treatment from a GO and receipt of SOC) continue to be critical points of intervention for improving survival time and reducing deaths from OC. Although it is difficult to determine when adjuvant chemotherapy is warranted based on sound clinical judgement (i.e., taking into consideration the patient’s comorbidities, toxicities, age, etc.) or patient refusal, one area that has not been carefully examined is the potential that race/ethnicity-based differences in patient and caregiver preferences may have for OC care. Future research may further explore this and the interaction effects of race, age, comorbidities on survival. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Our study showed that GOs more often provided the surgical and chemotherapeutic SOC. The receipt of surgical standards was associated with better survival outcomes, even after adjusting for provider specialty. As such, these two NCCN-recommendations (i.e., treatment from a GO and receipt of SOC) continue to be critical points of intervention for improving survival time and reducing deaths from OC. Although it is difficult to determine when adjuvant chemotherapy is warranted based on sound clinical judgement (i.e., taking into consideration the patient’s comorbidities, toxicities, age, etc.) or patient refusal, one area that has not been carefully examined is the potential that race/ethnicity-based differences in patient and caregiver preferences may have for OC care. Future research may further explore this and the interaction effects of race, age, comorbidities on survival. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. Supported by The United States Federal Government, Centers for Disease Control and Prevention, Atlanta, GA, United States.

Our study showed that GOs more often provided the surgical and chemotherapeutic SOC. The receipt of surgical standards was associated with better survival outcomes, even after adjusting for provider specialty. As such, these two NCCN-recommendations (i.e., treatment from a GO and receipt of SOC) continue to be critical points of intervention for improving survival time and reducing deaths from OC. Although it is difficult to determine when adjuvant chemotherapy is warranted based on sound clinical judgement (i.e., taking into consideration the patient’s comorbidities, toxicities, age, etc.) or patient refusal, one area that has not been carefully examined is the potential that race/ethnicity-based differences in patient and caregiver preferences may have for OC care. Future research may further explore this and the interaction effects of race, age, comorbidities on survival. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. Supported by The United States Federal Government, Centers for Disease Control and Prevention, Atlanta, GA, United States. Author contributions: Rim SH, Thomas CC and Stewart SL conceived of the study; Rim SH, Hirsch S, Thomas CC, Brewster WR, Cooney D, Thompson TD and Stewart SL wrote the paper and made substantial contributions to the conception, design, data analysis and interpretation of the data; Rim SH, Hirsch S, Thomas CC, Brewster WR and Stewart SL drafted and/or critically reviewed the manuscript for important intellectual content; all authors read and approved the final manuscript.

SL wrote the paper and made substantial contributions to the conception, design, data analysis and interpretation of the data; Rim SH, Hirsch S, Thomas CC, Brewster WR and Stewart SL drafted and/or critically reviewed the manuscript for important intellectual content; all authors read and approved the final manuscript. Institutional review board statement: The study was reviewed and approved by the Centers for Disease Control and Prevention Institutional Review Board through an expedited review process in accordance with standard procedures. Informed consent statement: The Institutional Review Board determined that informed consent was not needed, since the analytic dataset used contained only de-identified patient information and there was no patient contact in this study. Conflict-of-interest statement: The authors declare that they have no competing interests. Data sharing statement: No additional data are available. Core tip: A significant survival advantage is associated with receiving surgical standard of care (SOC), yet still some women had lower odds of receiving surgical SOC.

Informed consent statement: The Institutional Review Board determined that informed consent was not needed, since the analytic dataset used contained only de-identified patient information and there was no patient contact in this study. Conflict-of-interest statement: The authors declare that they have no competing interests. Data sharing statement: No additional data are available. Core tip: A significant survival advantage is associated with receiving surgical standard of care (SOC), yet still some women had lower odds of receiving surgical SOC. Figure 1 Surgical standard of care (n = 4434) and adjuvant chemotherapy standard of care (n = 2595) receipt by physician specialty and International Federation of Gynecologists and Obstetricians stage (1) Surgery SOC treatment was based on ovarian cancer patients receiving surgery prior to chemotherapy (n = 6714); (2) Stages 1, not otherwise specified and Unknown/unstaged were removed from analysis; (3) Surgeon specialty and chemotherapy specialty was categorized according to the most specialized care received during the course of the treatment window; (4) Women who received surgery SOC by a surgeon specialty who could not be identified are not shown (n = 17); (5) There were 177 women who received a chemotherapy procedure code of interest but for whom physician specialty could not be identified and 1238 women who did not receive a chemotherapy procedure code of interest.1 Denote that the estimate is statistically significantly higher for GO compared to Non-GO. SOC: Standard of care; GO: Gynecologic oncologist.

d a chemotherapy procedure code of interest but for whom physician specialty could not be identified and 1238 women who did not receive a chemotherapy procedure code of interest.1 Denote that the estimate is statistically significantly higher for GO compared to Non-GO. SOC: Standard of care; GO: Gynecologic oncologist. Figure 2 Ovarian cancer survivor curves1 by receipt of overall standard of care2 (n = 1678) 1All covariates held at the reference level noted in Table 3; 20 = Did not receive overall standard of care; 1 = Did receive overall SOC. SOC: Standard of care. Table 1 Characteristics of ovarian cancer patients who received any initial surgical procedure by physician specialty (n = 6714) Characteristic Surgeon specialty1 GO Non-GO

Figure 2 Ovarian cancer survivor curves1 by receipt of overall standard of care2 (n = 1678) 1All covariates held at the reference level noted in Table 3; 20 = Did not receive overall standard of care; 1 = Did receive overall SOC. SOC: Standard of care. Table 1 Characteristics of ovarian cancer patients who received any initial surgical procedure by physician specialty (n = 6714) Characteristic Surgeon specialty1 GO Non-GO OBGYN General surgeon Other2 No. of patients 2254 3088 914 419 Mean age at diagnosis (stddev) 74.6 (5.9) 74.8 (6.1) 77.0 (6.8)     75.5 (6.2) Race n (%) White 1995 (88.5) 2844 (92.1) 827 (90.5) 379 (90.5) African American 121 (5.4) 104 (3.4) 49 (5.4) 26 (6.2) Hispanic   35 (1.6)   31 (1.0) 3 3 Asian   53 (2.4)   66 (2.1) 3 3 Other4   47 (2.1)   37 (1.2) 3 3 Marital status Married 1052 (46.7) 1424 (46.1) 327 (35.8) 170 (40.6) Single 159 (7.1) 221 (7.2) 53 (5.8) 31 (7.4) Divorced 148 (6.6) 166 (5.4) 58 (6.3) 29 (6.9) Widowed   799 (35.4) 1168 (37.8) 458 (50.1) 176 (42.0) Separated/unknown   96 (4.2) 109 (3.5) 3 3 Charlson-Klabunde comorbidity score 0 1521 (67.5) 2133 (69.1) 605 (66.2) 266 (63.5) 1   498 (22.1)   644 (20.9) 188 (20.6)   93 (22.2) 2 175 (7.8) 189 (6.1) 78 (8.5) 38 (9.1) 3   45 (2.0)   80 (2.6) 29 (3.2) 3 4 or more 3   42 (1.4) 3 3 FIGO treatment stage IA/IB 200 (8.9)   383 (12.4) 66 (7.2)   43 (10.3) IC/II   276 (12.2)   516 (16.7) 90 (9.8) 40 (9.5) IIIA/IIIB 119 (5.3) 179 (5.8) 59 (6.5) 3 IIIC/IV 1580 (70.1) 1898 (61.5) 660 (72.2) 308 (73.5) Unstaged/NOS   79 (3.5) 112 (3.7) 39 (4.2) 3 Histology Serous 1460 (64.8) 1897 (61.4) 554 (60.6) 254 (60.6) Endometrioid   238 (10.6)   381 (12.3) 73 (8.0)   46 (11.0) Mucinous 129 (5.7) 235 (7.6) 79 (8.6) 25 (6.0) Clear cell   84 (3.7) 127 (4.1) 3 3 Adenocarcinoma   275 (12.2)   344 (11.1) 175 (19.1)   66 (15.8) Other5   68 (3.1) 104 (3.3) 20 (2.2) 3 1 Surgeon specialty was categorized according to the most specialized care received during the course of the treatment window

6 (11.0) Mucinous 129 (5.7) 235 (7.6) 79 (8.6) 25 (6.0) Clear cell   84 (3.7) 127 (4.1) 3 3 Adenocarcinoma   275 (12.2)   344 (11.1) 175 (19.1)   66 (15.8) Other5   68 (3.1) 104 (3.3) 20 (2.2) 3 1 Surgeon specialty was categorized according to the most specialized care received during the course of the treatment window 2 39 women received a surgery procedure code during the treatment window (defined as a period of two months prior and one year after a patient’s diagnosis date in which procedures were performed) but surgeon specialty could not be identified 3 Denotes cell size suppression of less than 20 4 Other race includes designation of “Other” or Native American 5 Other histology includes Transitional. GO: Gynecologic oncologists; FIGO: International Federation of Gynecologists and Obstetricians; NOS: Not otherwise specified. Table 2 Predictors of receipt of minimum surgical and chemotherapeutic standard of care1 Surgical standard of care2 Chemotherapeutic standard of care2

4 Other race includes designation of “Other” or Native American 5 Other histology includes Transitional. GO: Gynecologic oncologists; FIGO: International Federation of Gynecologists and Obstetricians; NOS: Not otherwise specified. Table 2 Predictors of receipt of minimum surgical and chemotherapeutic standard of care1 Surgical standard of care2 Chemotherapeutic standard of care2 Odds ratio (95%CI) P value Odds ratio (95%CI) P value Physician specialty3 Gynecologic oncologist 2.35 (2.03–2.71) < 0.01 1.25 (1.07–1.47)    0.006 Non-gynecologic oncologist 1.00 1.00 Age at diagnosis 66–69 1.00 1.00 70–74 0.80 (0.67–0.96)    0.017 0.93 (0.78–1.09)    0.393 75–79 0.83 (0.69–1.0)      0.053 0.79 (0.66–0.94)    0.008 80–84 0.58 (0.47–0.71) < 0.01 0.61 (0.48–75)   < 0.001 ⩾ 85 0.40 (0.31–0.51) < 0.01 0.31 (0.21–0.48) < 0.001 Race4 White 1.00 African American 0.67 (0.50–0.91)    0.01 – Other 0.83 (0.62–1.10)    0.208 – Treatment stage5 IA/IB 0.08 (0.07–0.10) < 0.01 NA    NA IC/II 0.08 (0.07–0.10) < 0.01 3.46 (2.86–4.18) < 0.001 IIIA/IIIB 0.05 (0.04–0.07) < 0.01 0.83 (0.64–1.09)    0.182 IIIC/IV 1.00 1.00 Charlson-Klabunde comorbidity score 0 1.00 1.00 1 0.84 (0.72–0.98)    0.029 0.84 (0.71–0.99)    0.029 2 0.81 (0.63–1.02)    0.084 0.78 (0.60–1.03)    0.078 3 0.65 (0.44–0.97)    0.039 0.49 (0.31–0.80)    0.005 4 or more 1.09 (0.60–1.97)    0.771 0.63 (0.29–1.37)    0.247 Histology Serous 1.00 1.00 Endometrioid 1.10 (0.90–1.35)    0.356 0.70 (0.56–0.89)    0.003 Mucinous 0.95 (0.74–1.35)    0.67 0.49 (0.34–0.70) < 0.001 Clear cell 1.29 (0.93–1.78)    0.13 0.62 (0.41–0.93)    0.026 Transitional 0.70 (0.27–1.79)    0.454 0.76 (0.30–1.97)    0.572 Adenocarcinoma (NOS) 0.44 (0.37–0.54) < 0.001 1.04 (0.86–1.27)    0.695 Other 1.05 (0.70–1.56)    0.813 0.74 (0.47–1.13)    0.168 Marital status Married 1.00 1.00 Not married 0.83 (0.72–0.95)    0.007 0.75 (0.66–0.86) < 0.001 Unknown 1.03 (0.69–1.52)    0.87 0.73 (0.48–1.09)    0.127 Year of diagnosis 1993–1997 0.62 (0.52–0.73) < 0.01 0.28 (0.23–0.33) < 0.001 1998–2002 0.79 (0.68–0.92)    0.003 1.09 (0.94–1.26)    0.261 2003–2006 1.00 1.00 SEER region4 Northeast – 1.00 Midwest – 0.76 (0.62–0.93)    0.009 South – 1.09 (0.88–1.37)    0.424 West – 0.93 (0.78–1.10)    0.391 1 Minimum SOC treatment was based on patients receiving surgery prior to chemotherapy (n = 6714)

< 0.001 1998–2002 0.79 (0.68–0.92)    0.003 1.09 (0.94–1.26)    0.261 2003–2006 1.00 1.00 SEER region4 Northeast – 1.00 Midwest – 0.76 (0.62–0.93)    0.009 South – 1.09 (0.88–1.37)    0.424 West – 0.93 (0.78–1.10)    0.391 1 Minimum SOC treatment was based on patients receiving surgery prior to chemotherapy (n = 6714) 2 Surgery SOC (n = 4434) and chemotherapy SOC (n = 2595) 3 Physician specialty was categorized according to the most specialized care received during the course of the treatment window; there were 39 and 177 cases where physician specialty could not be identified for surgery or chemotherapy procedures, respectively (results for this group not shown) 4 Race was not entered into the chemotherapy SOC model based on forward selection entry criteria (P ≤ 0.10); Region was not entered into the surgery SOC model based on forward selection entry criteria (P ≤ 0.10) 5 Stage I NOS, Stage IA/IB (for chemotherapy SOC) and unknown/unstaged were removed from the analysis since current guidelines recommend early stage patients not receive chemotherapy treatment. NOS: Not otherwise specified; SOC: Standard of care; SEER: Surveillance, Epidemiology, and End Result; NA: Not applicable. Table 3 Cox proportional hazard model of time-to-death among ovarian cancer patients Predictor Model 11 Model 21

5 Stage I NOS, Stage IA/IB (for chemotherapy SOC) and unknown/unstaged were removed from the analysis since current guidelines recommend early stage patients not receive chemotherapy treatment. NOS: Not otherwise specified; SOC: Standard of care; SEER: Surveillance, Epidemiology, and End Result; NA: Not applicable. Table 3 Cox proportional hazard model of time-to-death among ovarian cancer patients Predictor Model 11 Model 21 Hazard ratio (95%CI) P value Hazard ratio (95%CI) P value Received surgery SOC2 Yes 1.00 1.00 No 1.22 (1.12–1.33) < 0.01 1.21 (1.11–1.31) < 0.01 Received chemotherapy SOC2 4 Yes 1.00 4 No, but received some chemotherapy 0.95 (0.89–1.02) 0.18 4 Received no chemotherapy 1.29 (1.14–1.46) < 0.01 4 Age at diagnosis 66–69 1.00 1.00 70–74 1.07 (0.98–1.17) 0.13 1.05 (0.97–1.15) 0.24 75–79 1.23 (1.12–1.34) < 0.01 1.21 (1.10–1.32) < 0.01 80–84 1.52 (1.37–1.69) < 0.01 1.48 (1.33–1.65) < 0.01 ⩾ 85 1.96 (1.70–2.26) < 0.01 1.92 (1.67–2.21) < 0.01 Race White 1.00 1.00 African American 1.11 (0.95–1.29) 0.18 1.13 (0.97–1.32) 0.12 Other 0.90 (0.78–1.05) 0.17 0.88 (0.75–1.02) 0.09 Year of diagnosis 1993–1997 1.27 (1.17–1.38) < 0.01 1.24 (1.14–1.35) < 0.01 1998–2002 1.18 (1.09–1.27) < 0.01 1.17 (1.08–1.27) < 0.01 2003–2006 1.00 1.00 Treatment stage IA/IB 0.20 (0.18–0.23) < 0.01 0.17 (0.15–0.20) < 0.01 IC/II 0.35 (0.32–0.40) < 0.01 0.36 (0.32–0.40) < 0.01 IIIA/IIIB 0.61 (0.53–0.71) < 0.01 0.62 (0.54–0.71) < 0.01 IIIC/IV 1.00 1.00 Charlson-Klabunde comorbidity score 0 1.00 1.00 1 1.28 (1.18–1.38) < 0.01 1.26 (1.17–1.36) < 0.01 2 1.38 (1.22–1.56) < 0.01 1.37 (1.21–1.55) < 0.01 3 1.64 (1.34–2.00) < 0.01 1.64 (1.34–2.01) < 0.01 ⩾ 4 2.33 (1.73–3.15) < 0.01 2.27 (1.67–3.09) < 0.01 Histology Serous 1.00 1.00 Endometrioid 0.76 (0.68–0.85) < 0.01 0.75 (0.68–0.84) < 0.01 Mucinous 1.22 (1.06–1.41) < 0.01 1.22 (1.06–1.41) < 0.01 Clear cell 0.83 (0.69–1.00) 0.05 0.83 (0.69–1.00) 0.05 Transitional 0.79 (0.47–1.31) 0.36 0.79 (0.48–1.32) 0.37 Adenocarcinoma (NOS) 1.07 (0.98–1.18) 0.14 1.07 (0.97–1.17) 0.2 Other 1.02 (0.82–1.28) 0.85 1.02 (0.82–1.28) 0.85 Marital status Married 1.00 1.00 Not Married 1.07 (1.00–1.14) 0.05 1.07 (1.00–1.14) 0.05 Unknown 1.00 (0.82–1.23) 0.97 0.99 (0.80–1.21) 0.89 Surgeon specialty3 Non-GO 1.00 1.00 GO 0.90 (0.84–0.96) < 0.01 0.90 (0.84–0.97) < 0.01 Chemotherapy specialty3 Non-GO 4 1.00 GO 4 0.98 (0.89–1.08) 0.68 Did not receive chemotherapy 4 1.33 (1.19–1.47) < 0.01 1 Model 1 and Model 2: Includes OC patients who did not have an unknown FIGO stage at diagnosis, and survived at least 4.5 mo after diagnosis;

0 1.00 GO 0.90 (0.84–0.96) < 0.01 0.90 (0.84–0.97) < 0.01 Chemotherapy specialty3 Non-GO 4 1.00 GO 4 0.98 (0.89–1.08) 0.68 Did not receive chemotherapy 4 1.33 (1.19–1.47) < 0.01 1 Model 1 and Model 2: Includes OC patients who did not have an unknown FIGO stage at diagnosis, and survived at least 4.5 mo after diagnosis; 2 Minimum SOC procedure codes for surgery 3 Missing surgeon and physician specialty excluded from analysis 4 Excluded from the model based inclusion criteria. Chemotherapy SOC and chemotherapy physician specialty cannot be included in the same model because the common level of “did not receive chemotherapy” would introduce a singularity and prevent model convergence. OC: Ovarian cancer; SOC: Standard of care; GO: Gynecologic oncologist; FIGO: International Federation of Gynecologists and Obstetricians; NOS: Not otherwise specified. COMMENTS Background Ovarian cancer (OC) is the deadliest gynecologic cancer among women. Standard treatment for OC consists of extensive surgery and chemotherapy. Gynecologic oncologists (GOs) more often adhere to standard treatment guidelines among OC patients, resulting in longer patient survival. Research frontiers Low survival rates from OC may be related to lack of GO involvement in surgery or chemotherapy and/or lack of standard treatment receipt. Research measuring receipt of standard of care and its effect on survival can help reveal areas for intervention to improve OC mortality.

COMMENTS Background Ovarian cancer (OC) is the deadliest gynecologic cancer among women. Standard treatment for OC consists of extensive surgery and chemotherapy. Gynecologic oncologists (GOs) more often adhere to standard treatment guidelines among OC patients, resulting in longer patient survival. Research frontiers Low survival rates from OC may be related to lack of GO involvement in surgery or chemotherapy and/or lack of standard treatment receipt. Research measuring receipt of standard of care and its effect on survival can help reveal areas for intervention to improve OC mortality. Innovations and breakthroughs These results among over 6000 OC patients aged 65 and older indicate that a low proportion received standard treatment. Not having seen a GO, African American race, and being older (80+) were associated with not receiving standard treatment. Women who received standard treatment survived over one year longer than those who did not receive standard treatment. Applications Ensuring appropriate referral of OC patients to GOs for treatment will likely increase survival rates from OC. Education of primary care providers and/or health systems changes that promote referral would be beneficial to increase referral rates. Research is needed into patient factors and other potential reasons underlying lack of referral. Terminology GOs are subspecialists trained to administer both surgical and chemotherapeutic treatment to OC patients.

Applications Ensuring appropriate referral of OC patients to GOs for treatment will likely increase survival rates from OC. Education of primary care providers and/or health systems changes that promote referral would be beneficial to increase referral rates. Research is needed into patient factors and other potential reasons underlying lack of referral. Terminology GOs are subspecialists trained to administer both surgical and chemotherapeutic treatment to OC patients. Peer-review The authors investigated the influence of GO in the United States on surgical/chemotherapeutic SOC, and how this translates into improved survival among women with OC. The authors claimed that a survival advantage is associated with receiving surgical SOC and overall treatment by a GO. This manuscript provides useful information to the medical students, clinicians, and researchers in this field.