TCGA Cancer Clinical Data Resources provides a comprehensive --- the field of cancer research!

In 2006, the Cancer Genome Atlas (TCGA) program began a three-year pilot project with polyphystogram glioblastoma (GBM), pulmonary squamous cell carcinoma (LUSC) and ovarian fluid cystic adenocarcinoma (OV), which will be fully rolled out from 2009 to 2015.
the end of the project, TCGA network researchers have mas characterization of tumor molecules in more than 10,000 patients with 33 cancer types and defined many tumor molecular substypes.
TCGA contains clinically critical characteristics that represent generalized data collection.
to ensure the correct use of these genome-characterized large amounts of clinical data, TCGA recently completed a number of significant results published in the Journal of CellPress.
this article compiles an article published on Cell on April 5. the
article describes the standardized database of TCGA Pan-Cancer Clinical Data Resources (TCGA-CDR) with OS (total survival), PFI (no progression interval), DFI (disease-free interval), and DSS (disease-related survival) as the endpoints of the four main clinical outcomes, and gives recommendations for the use of endpoints for each cancer type.
TCGA clinical data can be downloaded from the Genomic Data Sharing Space (GDC) portal, and all molecular data is downloaded.
the same bar code structure for clinical and molecular data, facilitating the integration of clinical data and sample molecular data in patients.
TCGA Pan-Cancer Database Cohort Feature Map 1A is a flowchart for clinical data integration and analysis methods and four major clinical endpoint derivation and evaluation, and data on 33 initial registrations and 97 follow-up data files for 111,60 cancer patients of 33 cancer types were processed.
1 is the basic characteristic of each TCGA queue.
were selected into each queue according to molecular characteristics based on primary tumor samples, both primary and metastatic types of skin melanoma (SKCM), and very few other types of primary and metastasis tumors were also studied.
1. Clinical data integration and analysis methods and 4 major clinical endpoint derivation and evaluation flow chart 1. TCGA Pan-Cancer Queue Characteristics Clinical Outcome Endpoint OS, PFI, DFI, and Total Lifetime of DSS (OS) are important, with the advantage of defining OS events with minimal ambiguity.
but using OS as an endpoint may weaken clinical research, as non-cancer causes of death do not necessarily reflect tumor biology, invasiveness, or treatment remission.
use OS or take longer follow-up time;
for existing TCGA clinical data, it is important to recognize the importance of short clinical follow-up intervals in invasive cancer type outcomes, as clinical events may be observed within a few years and relapses or progressions occur before death.
for less aggressive types of cancer, patients relapse decades or even decades later, and enough events may not be observed during follow-up to support reliable results for judgment.
the purpose of this analysis is to examine the comparative advantages and deficiencies of TCGA's clinical efficacy test for pan-cancer, and to guide future analysis and avoid deficiencies such as insufficient follow-up interval.
analysis of all TCGA clinical data, the conclusion is as follows: OS, PFI and DFI are relatively accurate when data are available, but in most cases only DSS can be estimated.
1B is an OS K-M curve of 33 different types of cancer.
While survival analysis is not the primary goal of TCGA, most cancer type survival curves are essentially the same as independent studies that have evaluated the same end-of-life in the past, as evidenced by TCGA results for GBM, OV (2008, 2011) and low-grade glioma (LGG) (2015).
K-M curves of PFI, DFI and DSS are shown in Figure 1C-1E.
1 B-E clinical data analysis calculates the average follow-up time and the mid-term time of the event or examination based on the observation time of the 4 endpoints of each cancer type (Table 2).
The median follow-up time for all tumors was 22.1 months, but the time varied greatly between different types of cancer, with GBM and acute myeloid leukemia (LAML) as short as about 12 months and renal color cell carcinoma (KICH) at a maximum of about 48 months.
2 overall mid-level follow-up time and 4 endpoints of the event and observation of the clinical end point clinical end point recommended depends on the study objective, number of events, queue size, and result data quality.
these methods and other methods developed as tests and supplementary examinations of individual disease databases, giving recommendations and reasons for using the endpoint of each result in each disease type (Table 3).
end of each type of cancer requires at least one major test and supplemental examination to be accepted.
13 of the 33 cancers used all 4 endpoints: bladder urinary tract skin cancer (BLCA), cervical squamous cell carcinoma (CESC), colon cancer (COAD), esophageal cancer (ESCA), head and neck squamous cell carcinoma (ESCA) HNSC), renal papilloma cell carcinoma (KIRP), pulmonary adenocarcinoma (LUAD), LUSC, OV, pancreatic cancer (PAAD), sarcoma (SARC), gastric adenocarcinoma (STAD), and endometrial cancer (UCEC).
, none of the end points were chromoblastoma and sub-neurodegeneroma (PCPG).
lymphoma has only one endpoint for large B-cell lymphoma (DLBC), LAML, and thymus tumor (THYM);
most reliable is PFI, which can be unreservedly recommended for 4 tumors except LAML (no data), DLBC and KICH (cautious use) and PCPG (not recommended).
3 End point OS, PFI, DFI, and DSS assessment and recommendation using validated TCGA-CDR and case applications in breast cancer studies, estrogen ER-negative patients had worse clinical survival prognostication than those with ER-plus tumors.
used OS, PFI, DFI, and DSS to compare the lifetimes of these two types of tumor patients (Figure 3A-3D; curves up to 10 years follow-up time, but analyzed using the entire database).
single-factor analysis showed that patients with ER-plus breast cancer in TCGA had better survival than ER-patients using PFI (p-0.005) and DFI (p-0.001) as clinical endpoints, but there was no difference between the two types of patients when using OS (p-0.097).
addition, there were significant differences in DSS (p - 0.009) between the two groups of patients, indicating the potential value of DSS.
findings confirm that PFI and DFI are suitable endpoints for molecular studies of specific types of breast cancer.
3 validation and application instances also validate the endpoint of invasive GBM survival results.
the median GBM OS in TCGA was 12.6 months, between 12.1 months of the previously reported standard treatment and 14.6 months of the standard treatment combined with tymosamine.
median PFI was 6.1 months, between the previously reported 5 months of standard treatment and 6.9 months of standard treatment combined thymosamine.
, it is clear that the TCGA database OS and PFI event times are consistent with the literature.
once again confirmed the clinical endpoint effectiveness of OS and PFI as GBM molecular studies.
used cox proportional risk regression models to determine the risk ratio (HR) of cancer patients with relatively low periods (phases I, II) of the high period (phases III, IV) and to validate TCGA-CDR data for the four endpoints.
14 types of cancers recommended for OS, PFI, and DSS are compared to logHR (Table 3) because the definition of DFI is inconsistent with the definition of other results.
then only diseases that meet Cox scale risk assumptions are counted (Figure 3E-3G).
results showed that in 14 types of cancer, in addition to mesothelioma (MESO), PAAD and vine melanoma (UVM), high-period vs. low-stage HR was significantly higher in the three recommended endpoints OS, PFI and DSS.
Wilcoxon rank and test using paired samples, when measured using PFI and DSS (p s 0.0008) or PFI and OS (p s 0.039), there was a significant difference in logHR, showing a systematic deviation in disease progression and survival endpoint on HR;
to reverse-weighted average determination of Pearson correlation coefficients for standard errors of 2 logHR values, The three results estimated by logHR were highly positive: PFI and OS correlation coefficients of 0.96 (95% confidence interval: 0.77-0.99), PFI and DSS were 0.99 95 (95% confidence interval: 0.76-0.99), OS and DSS 0.90 (95% confidence interval: 0.61-0.98).
these correlations potentially support the use of PFI in the early stages of clinical use and OS and DSS in the later stages.
to integrate molecular data, the study also attempted to analyze whether different new tumor events occurred in patients who were disease-free after the first treatment and non-disease-free patients.
29 types of cancer information in TCGA-CDR are used to address the problem, including patients who survive for at least 3 months from diagnosis to completion of the first treatment and to a disease-free state.
example, there were 289 cases of non-patients and 41 cases of long-term non-disease-free patients, with NTE rates of 21.8% and 68.2%, respectively.
Using the Cox scale risk regression model, it was found that the risk of NTE was significantly higher in non-disease-free patients than in non-diseased patients (HR s 6.68, 95% CI s 4.25-10.51, FDR adjusted q value (0.05).
similar results in 21 other cancer types (Table 4).
were not observed in the remaining seven tumors.
we are also evaluating whether each model meets the scale risk hypothesis, two of which do not, and need to study time-dependent and multivariate models to find out why.
4 Non-disease-free vs. disease-free NTE Development Comparison TCGA collects cases from hundreds of sites around the world, taking into account data integrity, tumor and patient characteristics to address comparability between different sites of clinical data.
for each disease, we compared the top two tissue source sites with the highest number of cases (TSS) with all other TSS diseases and 4 end points (Table 5).
Table 5 compares the results of each type of cancer study from the first two TSS for highly invasive tumors such as #1 GBM, with the first two TSS patients making OS, PFI, and DSS similar to other TSS, because too few events are difficult to analyze and do not recommend DFI as an endpoint.
invasive tumors such as BRCA, PFI and DFI are highly recommended, but OS or DSS assessments are recommended with caution.
TSS #1OSDSS, but no PFI differences were observed, and on the surface, TSS1 had better DFI results, with only 3 DFI events.
in this table indicates that the site's results data need to be further evaluated.
the other hand, TSS #24.
comparison of the results suggests that specific information about TSS needs to be taken into account when analyzing clinical data for the overall TCGA from specific results.
TSS can be used as an alternative to other untested differences, such as these and incomplete clinical annotations, because factors such as age, tumor stage/grading, and treatment affect the endpoint outcomes of different TSS patients.
use of all clinical data treated for clinical end-of-life analysis based on potential mixing factors, risk of competitive outcomes, and model assumptions.
but compared to the integrated analysis of molecular/genomic data, we should pay attention to two aspects.
first, the recommendation is based on the baseline survival model, with molecular sub-types as predictors to divide the sample set, which may destroy the statistical significant differences in results.
therefore, conclusions drawn from interrelated TCGA molecular data or tumor subsystic TCGA-CDR results data need to be further confirmed in a separate tumor database.
second, we recommend using only molecular data for primary tumors, as matching clinical data such as important time information are collected at the time of initial diagnosis.
skin melanoma (SKCM) is very special in TCGA tumor types, with only 103 primary tumors in 470 cases, the remaining 296 cases as local lymph node metastasis and 68 remote metastasis.
contrasts with other TCGA cancer types that rarely collect metastatoma.
SKCM metastatomas rarely have matching primary tumors, while other TCGA cancer types, although rarely metastatomas, have matching primary samples.
therefore, for the relevance of SKCM results, it is recommended to use only a limited number of primary cases, although phase III cases of SCKM lymph node metastasis can be studied as separate groups.
the use of the newly consolidated TCGA-CDR database, it is also important to note: potential confusion factors, risk of competitive outcomes, and model assumptions.
a mix of factors but is excluded from the model, deviations may overestimate or underestimate the true efficacy.
, for example, in studies of racial differences in breast cancer, there were significant differences in gene expression between white and black patients, but when adjusted for molecular substations, the differences decreased significantly or even disappeared.
effects are also potentially mixed factors and adjustments should be given due consideration when information is available.
treatment is unknown, standard treatment such as age, diagnosis hospital and diagnosis year can reduce some errors as an alternative information.
model decisions in this area and encourage the use of tumor markers for prognosm research recommendation reports (REMARK).
DSS, DFI, and PFI endpoints for competitive results, patients who have not experienced significant events and who have not died of disease are reviewed.
case, if the patient is assumed to have no other cause of death, she/he may eventually die of cancer.