If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Does the primary site really matter? Profiling mucinous ovarian cancers of uncertain primary origin (MO-CUP) to personalise treatment and inform the design of clinical trials
Molecular Oncology Group, School of Women's and Children's Health, Faculty of Medicine, UNSW Sydney, NSW, AustraliaPrince of Wales Clinical School, Faculty of Medicine, UNSW Sydney, NSW, Australia
Department of Gynaecological Oncology, Westmead Hospital and Centre for Cancer Research, The Westmead Institute for Medical Research, The University of Sydney, NSW, Australia
Molecular Oncology Group, School of Women's and Children's Health, Faculty of Medicine, UNSW Sydney, NSW, AustraliaThe Kinghorn Cancer Centre, Garvan Institute of Medical Research, 384 Victoria Street, Darlinghurst, NSW 2010, Australia
Prince of Wales Clinical School, Faculty of Medicine, UNSW Sydney, NSW, AustraliaDepartment of Medical Oncology, Prince of Wales Hospital, Randwick, NSW, Australia
Advanced mucinous ovarian cancers are rare with few histotype specific trials.
•
Includes mucinous ovarian cancers of uncertain primary (MOCUP).
•
Tumor profiling MOCUPs to identify treatment targets for basket trials needed.
•
Largest reported series of MOCUPS and includes almost 300 invasive cancers.
•
Most common mutations were KRAS (60%), TP53 (38%), PIK3CA (13%) and PTEN (9%).
Abstract
Objective
Advanced stage mucinous ovarian cancers are diagnostically and therapeutically challenging. Histotype specific trials have failed due to low recruitment after excluding non-ovarian primaries. Mucinous ovarian cancers are commonly metastatic from other sites however lack definitive diagnostic markers. We suggest a classification of mucinous ovarian cancers of uncertain primary origin ‘MO-CUPs’ in clinical trials. This study aims to identify drug targets to guide treatment and future trials.
Methods
We analyzed a large de-identified, multi-platform tumor profiling dataset of MO-CUPs enriched for advanced stage and recurrent cases submitted to Caris Life Sciences. Available data included a 45-gene next-generation sequencing (NGS) panel, gene amplification of HER2 and cMET and 18 immunohistochemical (IHC) markers of drug sensitivity/resistance.
Results
Mucinous tumors from 333 patients were analyzed, including 38 borderline tumors and 295 invasive cancers. The most common mutations in a subset (n = 128) of invasive cancers were KRAS (60%), TP53 (38%), PIK3CA (13%) and PTEN (9%). Borderline tumors had higher rates of BRAF mutations, and PGP and TOP2A overexpression than invasive cases. KRAS mutant invasive cancers had lower expression of thymidylate synthase (p = 0.01) and higher expression of TUBB3 (p = 0.01) than KRAS wildtype tumors.
Conclusions
To our knowledge, this is the largest series profiling mucinous ovarian cancers and almost certainly includes cases of ovarian and non-ovarian origin. Given the difficulty recruiting patients to histotype-specific trials in rare subsets of ovarian cancer, it may be more important to focus on identifying potential treatment targets and to personalise treatment and design clinical trials in MO-CUPS agnostic of primary site to overcome these issues.
] and have a very good prognosis with surgery alone. In contrast, patients presenting with advanced stage mucinous cancers of presumed ovarian origin have a very poor prognosis and low response rates with carboplatin and paclitaxel, which are widely used to treat epithelial ovarian cancers of all histological subtypes [
] of all epithelial ovarian cancers (EOC) in historical series. More contemporary, and reliable estimates indicate that they are rare and make up only 2–4% of EOC [
Targeted deep sequencing of mucinous ovarian tumors reveals multiple overlapping RAS-pathway activating mutations in borderline and cancerous neoplasms.
]. The wide variability in incidence relates to the fact that mucinous tumors involving the ovary are often misclassified as ovarian in origin when they are more likely to be metastatic to the ovary. [
] The possible primary sites that metastasize to the ovary include the gastrointestinal tract (~45%), pancreas (20%), cervix/endometrium (18%), breast (8%), with the remainder unknown (4%) [
There have been several diagnostic algorithms proposed to help distinguish primary mEOC from metastatic mucinous cancers from other sites, including clinico-pathological factors such as uni- vs bi-laterality, tumor size [
]. Despite this, it remains challenging to confidently identify cohorts of patients with “true” primary advanced stage mEOC for analysis of treatment outcomes as well as for inclusion in clinical trials. Indeed, more effort has gone into trying to identify the primary site than improve the outcomes of patients with advanced mEOC.
Advanced stage mucinous cancers of the ovary have a much lower response to standard chemotherapy regimens used for EOC and a significantly worse survival compared to patients with other EOC subtypes [
]. A meta-analysis of seven randomised ovarian cancer trials found a median overall survival of only 14.6 months in advanced mEOC compared with a median survival of 40.8 months in patients with serous carcinoma [
The relative rarity of mucinous ovarian cancers and the inherent problems with misclassification mean that they are underrepresented in randomised clinical trials [
]. In late 2009, the first mucinous-specific trial (GOG 0241) opened with the optimistic aim of recruiting large numbers of patients with advanced or recurrent ‘mucinous ovarian cancers’. Patients were randomised to one of four arms; oxaliplatin and capecitabine, which is a commonly used regimen in colorectal cancer, or to “standard” ovarian cancer chemotherapy (carboplatin and paclitaxel), along with addition of bevacizumab or not with either regimen [
Multicentre trial of carboplatin/paclitaxel versus oxaliplatin/capecitabine, each with/without bevacizumab, as first line chemotherapy for patients with mucinous epithelial ovarian cancer (mEOC).
]. The trial was closed due to poor accrual, largely related to the fact that more than half of the patients (57%) were reclassified as being metastatic to the ovary rather than ovarian in origin after expert pathology review, which was one of the criteria for entry to the study. [
] With the benefit of hindsight, it may have been better to recruit all eligible patients with advanced stage mucinous cancers involving the ovary irrespective of primary site. Coupled with tumor collection and translational research, this trial may have been pivotal in improving our approach to management of these patients and was a lost opportunity.
Molecular profiling of invasive mucinous tumors involving the ovary including both ovarian and non-ovarian primary sites may help determine the organ of origin, but more importantly could identify potential treatment targets to guide patient management. Arguably, given the diagnostic uncertainty and problems associated with confidently assigning the primary site, clinical trials should be more pragmatic and include all patients with advanced stage mucinous adenocarcinomas involving the ovary of unknown origin and we propose an umbrella term of ‘mucinous ovarian cancers of uncertain primary origin (MO-CUP)’. This would enable recruitment of much larger groups of patients to basket trials, whereby eligibility is based on molecular characteristics rather than a definitive site of origin, and ultimately could lead to improved outcomes in specific well-defined subsets of patients.
There have been relatively few studies reporting the molecular profiles of MO-CUPs to date, with three series containing sample sizes averaging 34 invasive cases, [
Targeted deep sequencing of mucinous ovarian tumors reveals multiple overlapping RAS-pathway activating mutations in borderline and cancerous neoplasms.
Targeted deep sequencing of mucinous ovarian tumors reveals multiple overlapping RAS-pathway activating mutations in borderline and cancerous neoplasms.
Targeted deep sequencing of mucinous ovarian tumors reveals multiple overlapping RAS-pathway activating mutations in borderline and cancerous neoplasms.
Targeted deep sequencing of mucinous ovarian tumors reveals multiple overlapping RAS-pathway activating mutations in borderline and cancerous neoplasms.
Targeted deep sequencing of mucinous ovarian tumors reveals multiple overlapping RAS-pathway activating mutations in borderline and cancerous neoplasms.
] have also been reported. Both overexpression of HER2 by immunohistochemistry and amplification are consistently reported in up to 20% of invasive mucinous cancers [
], and are often mutually exclusive to KRAS mutations.
We report the results of analyzing multiplatform tumor profiling data from a large cohort of almost 300 women with invasive mucinous tumors with either known or suspected ovarian origin. We anticipate that this cohort is enriched for advanced stage/recurrent disease given that one of the most common reasons for submitting tumor specimens for profiling is to help plan treatment. We cannot definitively conclude this, however based on the limited clinical and diagnostic data provided at least 34% of cases had advanced stage/recurrent disease, and this could be an underestimate. The overarching aim was to identify distinct molecular subsets of MO-CUPs and potential therapeutic targets/treatment strategies that could be tested in clinical trials and inform the management of patients.
2. Methods
2.1 Patient cohort and clinical data
The study cohort consisted of women referred internationally to Caris Life Sciences for molecular testing between late 2009 and early 2017 and included all women with the description of a mucinous tumor of known or suspected ovarian origin based on information provided by the referring clinician. Referrals are made to Caris Life Sciences to help clinicians select chemotherapy or identify targeted agents that might be of benefit to their patients. De-identified extracts of clinical history and pathology report information were reviewed to assign histological subtype, grade, and the site of tumor tissue sent for profiling. In all cases, either the pathology report or the clinical history indicated a suspicion/presumptive diagnosis of mucinous ovarian cancer. The Western Institutional Review Board, the IRB for Caris Life Sciences deemed the study exempt from additional patient consent since it used previously collect de-identified data.
2.2 Tumor profiling tests
Tumor samples were tested at the request of the referring physician. A combination of tests including immunohistochemistry (IHC), gene amplification by either fluorescent or chromogenic in situ hybridisation (ISH), and/or hotspot mutations by next-generation sequencing (NGS). The panel of tests developed over time, with NGS introduced from January 2013. The specific IHC tests included in the Caris Life Sciences panel were based on published studies of associations with a response to treatment in one or more cancer types and varied over time and with evolving evidence (S1). Where results were indeterminate, these were re-coded to missing, and equivocal IHC staining was coded as negative (except in cases with equivocal HER2 staining, where ISH confirmed HER2 amplification). In the NGS results, variants of unknown significance were assigned as mutation negative.
In addition, we examined the most common mutation frequencies in our cohort compared with publicly available data through the American Association of Cancer Research (AACR) initiative Genomics Evidence Neoplasia Information Exchange ‘GENIE’ [
]. A search was run in January 2018 to find the frequencies for key mutations and HER2 amplification in mucinous tumors of the ovary, colon, appendix and pancreas.
2.3 Statistical analysis
Due to changing test practice over time, different analyses were performed on the cohort before and after 2013, when NGS data were available. The full cohort included results from 18 IHC tests as well as HER2 amplification for up to 295 invasive cases and 38 borderline tumors, and a subset of these from 2013 onward of 128 invasive cases and 19 mucinous borderline tumors had 18 IHC results, HER2 and cMET amplification data, as well as NGS data for 45 genes. A sensitivity analysis was performed to assess any changes in frequency between time periods, before and after 2013 (S2) with a comparison of test positive proportions using the chi-square statistic. We also compared frequencies across borderline (n = 38) and invasive cases to assess differences between subgroups, as well as stratifying the cohort by KRAS mutation status, seen in high frequencies in mucinous tumors including borderline tumors.
A sensitivity analysis was performed to stratify MO-CUPs that were highly suspicious of a non-ovarian primary based on the limited clinical and pathological data. The classification was based on diagnostic description of widespread metastatic disease involving multiple organs, often without ovarian involvement but a clinical history including the ovary. This sensitivity analysis used clustering analysis and sought to assess any differences in expression or mutation profiles for those MO-CUPs most likely to be non-ovarian.
Data management and statistical analyses were performed using SAS Version 9.4, and clustering analyses were done using R Studio. Differences in frequencies between subgroups within the cohort and compared to GENIE data were calculated with the Fisher's exact test using SAS v9.4. For each subject, the different types of data (e.g. NGS, IHC) were linked based on their de-identified subject ID. Imputation of missing values was performed for the integrated analysis using mean imputation [
]. Missing values for each variable were replaced with mean value in conjunction with an extra variable that identifies whether a value is imputed or not to include it in the analysis. Integrated clustering analyses and associated plots were produced using RandomForest [
]. A random forest model was used to classify samples into three groups (mucinous borderline ovarian tumors, ovary, and non-ovary) and the top 20 genes that contributed to this classification based on mean accuracy decrease score were used to cluster samples and generate the waterfall plot. The hierarchical clustering was performed using the Complete Linkage and the Euclidean distance as the dissimilarity metric [
An initial query on the Caris Life Sciences database returned 437 women aged over 18 years categorised with a suggested diagnosis of an ovarian, fallopian tube or primary peritoneal tumor with any mucinous component. After review of the pathology data, 104 were excluded as having a diagnosis other than a mucinous borderline tumor or adenocarcinoma, the majority of these being other histologies that contained a minor mucinous component (S3). More than half of invasive cases had missing information on tumor grade, for those with data 64% were high grade and 36% low grade (S4).
The final dataset (Table 1) included 295 mucinous adenocarcinomas, and 38 mucinous borderline ovarian tumors (mBOT). Supplementary Table 5 (S5) shows the tumor tissue sites sampled for profiling. Table 2 shows the most common molecular changes in invasive tumors, the most frequent mutations in KRAS (60%), TP53 (38%), and PIK3CA (13%).
When we compared mutation frequencies in mBOT and invasive tumors, BRAF mutations were significantly higher in mBOT compared to invasive tumors (21% vs 3%, p = 0.0012 (Fig. 1a ), as well as differences in the proportion of cases with overexpression of PGP (84% vs 56%, p = 0.0027), and TOP2A (32% vs 59%, p = 0.0018, Fig. 1b). Amplification of HER2 was seen in 3% of mBOT and 11% of invasive cases and no cases of mBOT had cMET amplification compared to 2% of invasive cases (data not shown), not statistically significant.
Fig. 1a: Hotspot mutation frequencies in mucinous borderline and MO-CUPs
As the most common change, we also stratified invasive mucinous tumors with KRAS mutation data (n = 127, S6) and found more KRAS mutant cases had loss of expression of TS (75% vs 55%, p = 0.034), and that KRAS wildtype were associated with greater TUBB3 loss (61% vs 80%, p = 0.020).
3.1 Integrated analysis
We found no differences between borderline or invasive tumors, including stratification of those MO-CUPs suspicious for non-ovarian origin, based on visualization and clustering of the molecular features among all samples. Supplementary Fig. S7 shows waterfall plots with similar mutation profiles across groups, and S8 shows that clustering analyses did not correlate with invasiveness or suspected primary origin.
A comparison of frequencies for each assay with data before and after 2013 found no differences for the vast majority, but statistically significant differences in the proportion of cases with PTEN loss, or overexpression of ERCC1, or cMET (S2). This could be a random finding, or may reflect a bias within the cohort due to a change in assay for these proteins.
3.2 Comparison with publicly available data
We compared our cohort with publicly available tumor profiling data for invasive mucinous cancers of the ovary (n = 36), colorectum (n = 149), appendix (n = 92), pancreas (n = 18), as well as high-grade serous ovarian cancers (n = 687) from the AACR GENIE database (Fig. 2). KRAS mutations were the most common alterations across all mucinous tumors, regardless of primary site. In contrast, KRAS mutations were exceedingly rare in high-grade serous ovarian carcinomas. TP53 mutations, the dominant alteration found in high-grade serous ovarian carcinomas (87%), although comparably lower in mucinous carcinomas, were consistently frequent across all tissue types (20–60%).
Fig. 2Key mutation frequencies compared with data from AACR GENIE
The greatest variation was observed between the pattern of mutations in all mucinous cancers compared with high-grade serous ovarian cancer. Other key differences were seen in our cohort compared with mucinous colorectal cancers, with lower mutation rates in PIK3CA (13% vs 23%), SMAD4 (4% vs 18%), GNAS (5% vs 14%) and BRAF (3% vs 21%), and a higher frequency of HER2 amplification (11% vs 1%). Mutations in GNAS were significantly more prevalent across all gastrointestinal sites (14–67%) compared with mucinous ovarian tumors (5%).
4. Discussion
To our knowledge, this is the largest reported series of invasive mucinous tumors, with known or suspected involvement of the ovary, using comprehensive, multi-platform integrated analysis to identify potential therapeutic targets. Data for a subset of this cohort were published in conference proceedings from 2015 [
] however this update includes a more rigorous review of pathology data for confirmation of histological subtype.
The key findings were that despite many tests performed, approximately 25% of these MO-CUPs have a profile that may be amenable to treatment with current targeted therapies based on aberrant HER2 or a mutation in the mTOR pathway (PIK3CA/PTEN/AKT), however mutations in other genes such as BRAF may also prove to be potential treatment targets. We took a pragmatic approach to classification of these cases due to limited information, yet subgroup analysis of those with pathology data highly suspicious of a non-ovarian primary did not reveal significant molecular differences, suggesting that advanced stage mEOCs may have more in common with mucinous cancers arising in other sites beyond sharing a similar histological phenotype [
]. This was confirmed by our comparison with publicly available data of mucinous cancers from different primary sites where despite differences in some genes, the mutation patterns were consistent irrespective of tumor origin. These analyses also highlighted the stark difference between mucinous and high-grade serous ovarian cancers, suggesting that the practice of using the same chemotherapy treatment protocol across these ovarian histotypes may not be appropriate and given the low response rates with carboplatin and paclitaxel in mEOCs, [
] remains an important issue that must be addressed.
The major challenge here and in prior observational studies and trials is the difficulty in classifying mucinous cancers that involve the ovary with respect to likely primary site. This has resulted in the status quo being maintained, and there has been no appreciable progress in the management of patients with advanced stage mucinous ovarian tumors. This is well illustrated in the GOG 182 trial which included 54 mucinous tumors out of almost 4000 epithelial ovarian cancers. Pathology review by 3 pathologists suggested that only 16–18% were definitely ovarian in origin and the others judged to be metastatic to the ovary. However, there was no difference in median overall survival which was only 14 months in the 2 groups compared to 42 months in the non-mucinous EOC patients [
]. The prognosis is poor for patients with advanced stage mucinous cancers regardless of the primary site and we suggest that rather than focussing efforts on assigning the primary site with complex algorithms, it would be more important to identify potential therapeutic targets in women with advanced stage mucinous cancers, irrespective of the site of origin and include all eligible patients in clinical trials that include tumor collection and translational endpoints.
The frequencies of mutations in key genes such as KRAS and TP53 are consistent with prior studies. Our population has slightly lower levels of HER2 overexpression (13%) and amplification (11%) than other published series (18–19%), although not statistically significant upon comparison of proportions using the chi-squared statistic, and consistent with publicly available GENIE data (8%). The exception is Chay's report of a large Asian cohort (n = 113) of predominantly (80%) stage 1 disease that found HER2 positivity in 27% of cases. Of additional note, 89% (17/19) of borderline tumors in our series had a mutation in either KRAS or BRAF, confirming prior reports that this pathway activation is likely to be an early event in the development of these tumors [
]. The limited clinical data available indicates that this cohort represents a subset of patients with more advanced, aggressive, or recurrent mucinous tumors (reported for 34%) where clinicians requested tumor profiling to help inform treatment for their patients and this fact should be considered when comparing our results to prior research cohorts. Any differences in representativeness may be less marked in comparisons with GENIE data, drawn from U.S. tertiary referral hospitals performing in-house molecular profiling to plan treatment in-line with the Caris Life Sciences platform described here.
In the context of applying the findings in this study to plan treatment or design trials, there are some molecular and chemotherapy targets within this cohort that could be further explored. Loss of expression of RRM1, ERCC1 and TS have been reported to be associated with response to a range of chemotherapeutic agents in some cancer types including gemcitabine, cisplatin and 5-fluorouracil respectively [
]. It is relevant to note that a recently reported randomized phase 2 trial of carboplatin and paclitaxel ± trastuzumab in patients with HER2 positive serous carcinoma of the uterus, another rare gynecological cancer, found that there was a significant benefit associated with the addition of trastuzumab to chemotherapy [
Randomized phase II trial of carboplatin-paclitaxel versus carboplatin-paclitaxel-trastuzumab in uterine serous carcinomas that overexpress human epidermal growth factor receptor 2/neu.
]. In addition, it would be worth exploring whether the differences in KRAS mutation status could be used to predict sensitivity to 5-flurouracil (KRAS-mutant) vs taxane-based therapy (KRAS-wildtype) based on the TS and TUBB3 loss respectively. Expression levels of TUBB3 have been shown to be associated with response to taxane-based therapies [
The prognostic role of the class III β-tubulin in non-small cell lung cancer (NSCLC) patients receiving the taxane/vinorebine-based chemotherapy: a meta-analysis.
]. It will be important to validate these markers in larger cohorts, and ultimately design a trial for all women with MO-CUPs where these questions and many others could be addressed.
An important limitation of this study is that we did not evaluate data on microsatellite instability (MSI) which has been reported previously in 27% of mucinous colorectal cancers [
BRAF mutation, CpG island methylator phenotype and microsatellite instability occur more frequently and concordantly in mucinous than non-mucinous colorectal cancer.
]. In addition, almost 40% of this cohort had PD-1 overexpression but we did not have PD-L1 expression data. We also had no data on tumor mutational load or homologous recombination repair deficiency (HRD) as the patients included in our cohort had their tumors analyzed before all these parameters were included in the routine profiling panel. More recent estimates from Caris Life Sciences from 2017 indicates MSI positivity in approximately 5% (3/61) and PD-L1 expression in 8% (6/73) of cases of mucinous ovarian cancer (data not shown). Going forward, MSI and PD-L1 expression may identify a subset of patients who would benefit from PD-1/PD-L1 inhibitors and routine testing opens the possibility of more effective treatment in this subgroup of patients [
]. HRD may also help identify a subset that could respond to PARP inhibitors.
A further limitation is that we have incomplete clinical and pathological data, and the criteria for inclusion was the diagnosis of a mucinous cancer considered to be ovarian in origin by the referring clinician. In 102/295 (34%) of invasive cases, the clinical history or diagnostic data indicated metastatic or recurrent disease which is higher than the reported 20% of advanced stage mEOC at initial presentation reported in the literature [
]. The dataset almost certainly includes patients with non-ovarian primary sites based on the results of other studies and our review of the available pathology data, but nonetheless this reflects clinical reality where it is often very difficult to know the primary site. Not all cases had all tests performed, however this was overcome by restricting integrated analyses to the smaller dataset, and imputation of missing values for this reduced cohort. We acknowledge that there were differences in the expression levels by proportion for three proteins, PTEN, ERCC1 and cMET which may be a source of bias however these proteins do not form the basis of any major findings reported here. Even though this is one of the largest series reported, some analyses, in particular clustering by suspected origin may be underpowered to detect differences. We also acknowledge that this population consists of women who have undergone commercial tumor profile testing either through health insurance or through self-funding which should be considered when assessing the generalizability of these findings.
There is increasing uptake of multi-platform molecular profiling to help guide treatment options in oncology. For this population, no established trials of mucinous tumors involving the ovary exist, making it challenging to determine the utility of this profiling in the absence of treatment options. We were unable to follow up this cohort to determine whether profiling had an impact on their management and have no knowledge of any outcomes of treatment. This population represents the ‘real-world’ dilemma of large numbers of patients with rare/uncommon tumor types undergoing molecular profiling to help identify potential treatment targets and select therapy.
It is only by collating and integrating this information in larger numbers of cases that we can start to identify common targets that could be worth investigating in clinical trials. This study is limited by the lack of clinical data and central pathology review, missing data for some cases, and possible issues in representativeness since it was drawn from a commercial tumor profiling cohort. The important hypothesis generated from this observational study is the concept that knowing the primary site may not matter in patients with advanced stage mucinous cancers. Rather than focussing solely on immunohistochemical tests and imaging to try and determine the most likely primary site - we should either include all eligible patients in basket trials which include biopsies and translational endpoints or offer multiplatform molecular profiling to personalize treatment for individual patients with MO-CUPS when trials are not available and ideally collect and report the outcome of treatment. We encourage the adoption of the term MO-CUP, both clinically and in a research context to address some of the limitations this uncertainty can cause and pose the question – does the primary site really matter?
Funding
Author NSM is supported by the NSW Ministry of Health and UNSW Sydney under the NSW Health PhD Scholarship Program, and the Translational Cancer Research Network, a translational cancer research centre program funded by the Cancer Institute NSW (RG171797 and 15/TRC/1-03). MF is supported by an NHMRC Program Grant, ID: APP1092856.
Conflict of interest statement
Authors KS and AV are paid employees of Caris Life Sciences who provided the data for this study.
Targeted deep sequencing of mucinous ovarian tumors reveals multiple overlapping RAS-pathway activating mutations in borderline and cancerous neoplasms.
Multicentre trial of carboplatin/paclitaxel versus oxaliplatin/capecitabine, each with/without bevacizumab, as first line chemotherapy for patients with mucinous epithelial ovarian cancer (mEOC).
Randomized phase II trial of carboplatin-paclitaxel versus carboplatin-paclitaxel-trastuzumab in uterine serous carcinomas that overexpress human epidermal growth factor receptor 2/neu.
The prognostic role of the class III β-tubulin in non-small cell lung cancer (NSCLC) patients receiving the taxane/vinorebine-based chemotherapy: a meta-analysis.
BRAF mutation, CpG island methylator phenotype and microsatellite instability occur more frequently and concordantly in mucinous than non-mucinous colorectal cancer.
31077-1&id=gr1.jpg){kind=link}
31077-1&id=gr2.jpg){kind=link}