Cholangiocarcinoma.org Statistics (CONTAINS SURVIVAL STATISTICS)
- This topic has 95 replies, 14 voices, and was last updated 10 years, 9 months ago by
.
-
Posts
-
Hi, Bruce,
Thanks for your tremendous analytical skill. Extreme helpful.
I have just come home from ASCO; and I have to ask you to see whether one of the session presenter’s opinion is valid or not. I talked to him about the usage of the HR, the p value and the 95% CI with regard to determine ,compare and validate the studies of, for example, of the efficacy,OS and toxicities on three studies of the SAME drug combo. GEM/CIS.(assumed the 3 studies are very similar in population, age and risk factors, stages ,and metastasis.etc. )
Here is the interesting part ;he told me after the session when I approached him and asked him,” If you are sitting on the FDA committee and trying to approve or disapprove a new drug, what is the most important statistic tool (HR,Bruce, will you shed some light on his comment to help me to understand his logic? I ask how about the HR and p value? He said no, He will just use the CI?
God bless.
I have corrected the original post with the plot of survival for Resected SEER Intrahepatic CC Patients. Median survival increased from 2.50 years to 3.42 year for 0 positive node patients and 1.25 to 1.33 years for 1 positive node patients.
Percy – I tried to include 95% confidence intervals for the two curves, but had trouble. However, I do have 95% confidence intervals for the median survivals. For the 0 positive node case, the 95% confidence interval is 2.33 – 4.50 years and for the 1 positive node case, it is 1.17 – 1.67 years.
Bruce
Percy – There are a lot of explanations on the web (Wikipedia, You Tube, etc.) concerning the “95% confidence interval,” “null hypothesis,” and p-value, but they generally involve statistical terms and numbers and are not representative of survival analysis.
The survival curves I’ve plotted so far are based on samples of cholangiocarcinoma.org data and the SEER data. Both of these databases are themselves only a sample of all the cholangiocarcinoma cases in the cholangiocarcinoma universe, which would be the world for cc.org and the USA for the SEER data. The curves are all approximate, since I’m working with a sample and not the whole universe. If I used another sample of the same size, I would get a different curve. If I had 100 samples, each the same size, I would get 100 different curves. On average, 95 of these curves would lie between an upper curve and a lower curve. This is the 95% confidence interval. The R statistical software I use plots the 95% confidence interval for each curve and I will do that for my corrected SEER IHCC curves to better illustrate this. I haven’t shown them in the past because they clutter up the plot.
The “null hypothesis” assumes there is no relationship between two measured phenomena. For example, in my last plot, the “null hypothesis” assumes that survival will be be the same with 0 or 1 positive node. If this was the case, the curves would lie on top of each other, but they could be separated solely because they are two independent samples. The question, are they separated because they are just two different samples, or are they separated because survival is dependent on positive nodes. The answer is in the p-value, which gives the probabilitiy that they are different because they are just two independent samples. Generally researchers would use a p-value equal to or less than 0.05 as evidence that the difference is because survival is dependent on positive nodes (>=95%) and not just different samples (<=5%) Bruce
Bruce,
This is a fabulous stream of information. Thank you for your energy and determination in gathering all this information and making sense of it for the rest of us.
I will send you information on my wife to add to your database.
To all who have added comments; You are definitely a great support group and I want to say thank you as a caregiver. I look forward to reading the comments in this message stream.
Carl
Hi, Bruce,
Well done.
One request, can you recommend some article that I can read and understand about the ” 95%confident interval “
God bless.I couldn’t find surgical margin or chemo data in SEER so I emailed them about it. Their response, in short, was that they do not collect surgical margin data and the limited chemo data they collect are not released to the public but may be available through special request. Radiation data is in SEER.
I was hoping to use SEER data to determine the effectiveness of adjuvant therapy for cc, but I suppose I will have to depend on posts to these discussion boards to calculate this.
Bruce
Below are histology codes and malignant tumors found in SEER which are not intrahepatic cholangiocarcinoma but are considered as intrahepatic bile duct cancers.
8000/3: Neoplasm, malignant
8001/3: Tumor cells, malignant
8010/3: Carcinoma, NOS
8012/3: Large cell carcinoma, NOS
8020/3: Carcinoma, undifferentiated type, NOS
8031/3: Giant cell carcinoma
8032/3: Spindle cell carcinoma
8033/3: Pseudosarcomatous carcinoma
8041/3: Small cell carcinoma, NOS
8046/3: Non-small cell carcinoma
8050/3: Papillary carcinoma, NOS
8070/3: Squamous cell carcinoma, NOS
8124/3: Cloacogenic carcinoma
8140/3: Adenocarcinoma, NOS
8141/3: Scirrhous adenocarcinoma
8161/3: Bile duct cystadenocarcinoma
8180/3: Combined hepatocellular ca. & cholangiocarcinoma
8240/3: Carcinoid tumor, malignant
8246/3: Neuroendocrine carcinoma
8260/3: Papillary adenocarcinoma, NOS
8310/3: Clear cell adenocarcinoma, NOS
8440/3: Cystadenocarcinoma, NOS
8453/3: Intraductal papillary-mucinous carcinoma, invasive
8470/3: Mucinous cystadenocarcinoma, NOS
8480/3: Mucinous adenocarcinoma
8481/3: Mucin-producing adenocarcinoma
8490/3: Signet ring cell carcinoma
8500/3: Infiltrating duct carcinoma, NOS
8560/3: Adenosquamous carcinoma
8800/3: Sarcoma, NOS
8980/3: Carcinosarcoma, NOS
9100/3: Choriocarcinoma
9120/3: HemangiosarcomaBruce
Eli – Thanks for the information on the Klatskin tumors being incorrectly cross-references to intrahepatic cc. The articles were written in 2006 and 2007, so I would have thought that the problem had been corrected in the Collaborative Stage Data Collection System version 2:0204, dated 12/5/11, which I used. Unfortunately, it hasn’t been corrected. I hadn’t downloaded histology data for the intrahepatic bile duct cancer data, but I found it as one of the parameters available and downloaded it. It refers to ICD-0-3. There were 779 patients coded 8162/3 Klatskin and included in the intrahepatic bile duct data.
I also noticed that there were 2,200 other patients among the 11,789 I downloaded as intrahepatic bile duct cancer, which were not coded as cholangiocarcinoma. I’ll need to eliminate the Klatskin and these other patients from the analysis and revise the above plots. Evidently bile duct cancer and cholangiocarcinoma are not the same.
Bruce
Oh my, you guys are amazing!! I am afraid I am a statistical idiot for the most part. I have a research thesis to do this year and it is daunting to me.
I so admire your ability to make sense of things and explain it to those of us that are statistically-challenged.
-Randi-
Many many thanks for this Bruce. I know that it must take you quite some time to go through all of the data etc to produce these reports here. Thanks again for all of your work and efforts with this.
Best wishes,
Gavin
Bruce, take a look at the articles below. Klatskin (hilar) tumors were misclassified as intrahepatic in the second edition of the ICD-O. Misclassification affected SEER data.
Impact of classification of hilar cholangiocarcinomas (Klatskin tumors) on the incidence of intra- and extrahepatic cholangiocarcinoma in the United States.
http://www.ncbi.nlm.nih.gov/pubmed/16788161Re: Impact of Classification of Hilar Cholangiocarcinomas (Klatskin Tumors) on Incidence of Intra- and Extrahepatic Cholangiocarcinoma in the United States
http://jnci.oxfordjournals.org/content/99/5/407.1.longI haven’t done the extrahepatic CC cases yet, but I should be able to do those.
I was hoping to find SEER data to do the two components of extrahepatic cc separately. These two components are refererred to in the 7th edition of the AJCC as perihilar (aka Klatskin) and extrahepatic (aka distal).
CS Schema v0204 shows bile duct data as Bile Ducts Intrahepatic (11,789 patients), Bile Ducts Perihilar (13,099 patients) and Bile Ducts Distal (387 patients). Most of the Bile Ducts Distal data is from 2010, which was the year the 7th edition of the AJCC became effective and extrahepatic cc was divided into perihilar and distal. For 2010, bile duct data in CS Schema v0204 is divided into intrahepatic (881, 51.0%), perihilar (510, 29.5%), & distal (337, 19.5%). The distribution for extrahepatic cc is, perihilar (60.2%) and distal (39.8%). Using 39.8% for distal, I would have expected to see about 5,367 cases of distal bile duct cancer in CS Schema v0204, instead of the very small 387 cases that are there, so something is definitely wrong with the process in which pre-2010 cases of extrahepatic cc has been divided into perihilar and distal.
Bruce
good information Bruce. Maybe I haven’t read all of your posts, but have you done the stats for extrahepatic CC?
EDIT [The original post was made after the original survival analysis. The woman referred to actually had cystadenocarcinoma, not cholangiocarcinoma. The longest surviving cholangiocarcinoma patient was a San Francisco County, CA man who died 26.6 years after his diagnosis.]
I forgot to mention that the longest surviving patient in the CS Stema v0204 intrahepatic cc section is a woman, diagnosed in January, 1977 in Alameda County, CA, who was still alive on 12/31/2010, a 33.9 years survivor. She is not included in the analysis since there was no lymph node data for her.
Bruce
- The forum ‘General Discussion’ is closed to new topics and replies.