This is the nineteenth blog in a series of blogs on neoplasia.
Though there has been little success in curing the advanced* common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood. Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? In this blog, and in the next few blogs, we'll be exploring the properties of common tumors and how they differ from the properties of rare tumors.
As we saw yesterday, the general approach to funding cancer research is to concentrate funding on the common tumors (the most difficut tumors to treat).
Rare tumors are different from common tumors. Differences between rare and common tumors account for the successes in treating some of the rare tumors.
The rare tumors tend to arise from the inside cells of the body, not the surface cells (as per yesterday's post). Among the rare tumors are the tumors of infancy and childhood. Almost none of the tumors of infancy and childhood develop from surface cells. Pediatric tumors do not follow many years of exposure to a carcinogen.
Congenital neoplasms (rare tumors present at birth) have, at most, nine months to develop. This indicates that they do not have the long development phase that characterizes the common tumors. The long development phase of the common tumors contributes to the accumulation of the numerous genetic alterations that are found in all of the common cancers (and absent from the congenital neoplasms).
There are few causes for any particular rare tumor (that is why a rare tumor is rare). A known cause of a rare tumor is likely to be the cause of any example of the rare tumor. The cause might be an inherited mutation, as is the case for inherited retinoblastoma. The cause might be a single exposure to an identified carcinogen at a documented moment in time, as in gestational exposure to diethylstilbestrol resulting in clear cell carcinoma of the cervix in adolescents.
Rare tumors are likely to have a single cause, or a single aberrant pathway, or a sigle inherited gene. By attacking simple changes in the cells of a rare tumor, it might be possible to arrest the growth of the tumor. It is difficult to imagine that knocking out a growth pathway in a heterogeneous population of genetically diverse cells, as we find in the advanced common tumors, will result in a cancer cure.
-Copyright (C) 2008 Jules J. Berman
*An advanced cancer is one that has directly spread extensively from its primary site or that has metastasized to a distant site
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Thursday, July 31, 2008
Friday, July 25, 2008
Neoplasms: 18
This is the eighteenth blog in a series of blogs on neoplasia.
Though there has been little success in curing the advanced* common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood. Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? In this blog, and in the next few blogs, we'll be exploring the properties of common tumors and how they differ from the properties of rare tumors.
As we saw yesterday, the general approach to funding cancer research is the same now as it was in the early 1970s; attack cancer tumor by tumor, dividing available funds in rough proportion to the number of people who die from each tumor.
The only problem with this straightforward approach is that it has not worked very well. Despite decades of funding, We still do not know how to cure the most common advanced (with extensive spread or metastases) cancers occurring in humans.
The reason that we have done so poorly with common tumors relates to the manner in which the common tumors occur.
The human body can be envisioned as a topological donut (Figure below) . Like donuts, we have a continuous surface. We can think of our skin as the outer-edge surface of the donut. We can think of our alimentary tract as the surface of the donut that lines the donut hole. Our outer-edge surface is lined by the epithelial squamous cells of the skin's epidermis. Our inner-edge surface is lined by epithelial enterocytes of the gastrointestinal mucosa.
The donut pastry cells correspond to the non-surface cells of the human body, the cells that arise from human mesoderm, which produces the connective tissues of the body (fibrous tissue, adipose tissue, muscle, bone, vessels). The non-surface cells also include the nervous system and gonads.
Virtually all exposure to toxic and carcinogenic chemicals is via the donut surfaces (the skin, and gastrointestinal tract), and the epithelial organs that bud off these surfaces (lungs, pancreas, liver, breast, prostate). Since the human surfaces contain the cells that are most exposed to carcinogens, it's not surprising that most human cancers are tumors of the surface cells. Just two cancers arising from the outside surface cells (squamous carcinoma of skin and basal cell carcinoma of skin) account for over one million new tumors each year in the U.S., a number that is roughly equal to all the other tumors of the body combined. Tumors of the surfaces account for over 95% of the tumors that occur in humans.
In the next blog, we'll look at the common properties of advanced common cancers that account for their poor cure rate.
-Copyright (C) 2008 Jules J. Berman
*An advanced cancer is one that has directly spread extensively from its primary site or that has metastasized to a distant site
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Though there has been little success in curing the advanced* common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood. Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? In this blog, and in the next few blogs, we'll be exploring the properties of common tumors and how they differ from the properties of rare tumors.
As we saw yesterday, the general approach to funding cancer research is the same now as it was in the early 1970s; attack cancer tumor by tumor, dividing available funds in rough proportion to the number of people who die from each tumor.
The only problem with this straightforward approach is that it has not worked very well. Despite decades of funding, We still do not know how to cure the most common advanced (with extensive spread or metastases) cancers occurring in humans.
The reason that we have done so poorly with common tumors relates to the manner in which the common tumors occur.
The human body can be envisioned as a topological donut (Figure below) . Like donuts, we have a continuous surface. We can think of our skin as the outer-edge surface of the donut. We can think of our alimentary tract as the surface of the donut that lines the donut hole. Our outer-edge surface is lined by the epithelial squamous cells of the skin's epidermis. Our inner-edge surface is lined by epithelial enterocytes of the gastrointestinal mucosa.
The donut pastry cells correspond to the non-surface cells of the human body, the cells that arise from human mesoderm, which produces the connective tissues of the body (fibrous tissue, adipose tissue, muscle, bone, vessels). The non-surface cells also include the nervous system and gonads.
Virtually all exposure to toxic and carcinogenic chemicals is via the donut surfaces (the skin, and gastrointestinal tract), and the epithelial organs that bud off these surfaces (lungs, pancreas, liver, breast, prostate). Since the human surfaces contain the cells that are most exposed to carcinogens, it's not surprising that most human cancers are tumors of the surface cells. Just two cancers arising from the outside surface cells (squamous carcinoma of skin and basal cell carcinoma of skin) account for over one million new tumors each year in the U.S., a number that is roughly equal to all the other tumors of the body combined. Tumors of the surfaces account for over 95% of the tumors that occur in humans.
In the next blog, we'll look at the common properties of advanced common cancers that account for their poor cure rate.
-Copyright (C) 2008 Jules J. Berman
*An advanced cancer is one that has directly spread extensively from its primary site or that has metastasized to a distant site
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Thursday, July 24, 2008
Neoplasms: 17
This is the seventeenth blog in a series of blogs on neoplasia.
Though there has been little success in curing the advanced common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood. Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? In this blog, and in the next few blogs, we'll be exploring the properties of common tumors and how they differ from the properties of rare tumors.
First, lets look at common cancers and see how we've been approaching the problem of their cure.
The general approach to funding cancer research is the same now as it was in the early 1970s; attack cancer tumor by tumor, dividing available funds in rough proportion to the number of people who die from each tumor.
Yesterday, we looked at the top cancer killers of Americans.
List. U.S. death rates and trends for the top 15 cancer sites [1], expressed as rates per 100,000 and age-adjusted.
Let's compare this with the top funding for cancers, cancer-by-cancer.
List. Cancer funding by NCI, by cancer site, expressed as millions of dollars spent in 2007; data from Office of Budget and Finance [2].
All of the other sites receive much less funding. For example, stomach gets $12 million and uterine cancer gets $16.6 million.
Is it not reasonable to fund cancer site-by-site, proportional to the number of people dying from cancer at the different sites? Is there anything wrong with this approach?
This will be the subject of the next blog.
1. [Seer Cancer Stat Fact Sheets. Table I-26 Age-adjusted U.S. death rates and trends for the top 15 cancer sites]
2. [Funding for Various Research Areas.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Though there has been little success in curing the advanced common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood. Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? In this blog, and in the next few blogs, we'll be exploring the properties of common tumors and how they differ from the properties of rare tumors.
First, lets look at common cancers and see how we've been approaching the problem of their cure.
The general approach to funding cancer research is the same now as it was in the early 1970s; attack cancer tumor by tumor, dividing available funds in rough proportion to the number of people who die from each tumor.
Yesterday, we looked at the top cancer killers of Americans.
List. U.S. death rates and trends for the top 15 cancer sites [1], expressed as rates per 100,000 and age-adjusted.
All Sites 189.8
Lung and Bronchus 54.1
Colon and Rectum 18.8
Breast 14.1
Pancreas 10.6
Prostate 10.1
Leukemia 7.4
Non-Hodgkin Lymphoma 7.3
Let's compare this with the top funding for cancers, cancer-by-cancer.
List. Cancer funding by NCI, by cancer site, expressed as millions of dollars spent in 2007; data from Office of Budget and Finance [2].
Lung Cancer $226.9 million
Colorectal Cancer $258.4 million
Breast Cancer $572.4 million
Pancreatic Cancer $73.3 million
Prostate Cancer $296.1 million
Leukemia $205.5 million
Non Hodgkin's Lymphoma $113.0 million
All of the other sites receive much less funding. For example, stomach gets $12 million and uterine cancer gets $16.6 million.
Is it not reasonable to fund cancer site-by-site, proportional to the number of people dying from cancer at the different sites? Is there anything wrong with this approach?
This will be the subject of the next blog.
1. [Seer Cancer Stat Fact Sheets. Table I-26 Age-adjusted U.S. death rates and trends for the top 15 cancer sites]
2. [Funding for Various Research Areas.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Wednesday, July 23, 2008
Neoplasms: 16
This is the sixteenth blog in a series of blogs on neoplasia.
Though there has been little success in curing the advanced common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood. Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? In this blog, and in the next few blogs, we'll be exploring the properties of common tumors and how they differ from the properties of rare tumors.
First, lets look at common cancers and see how we've been approaching the problem of their cure.
The general approach to funding cancer research is the same now as it was in the early 1970s; attack cancer tumor by tumor, dividing available funds in rough proportion to the number of people who die from each tumor. Here are the top 15 cancer killers in the U.S. (Data from the National Cancer Institute's SEER, Surveillance Epidemiology and End Results Program).
List. U.S. death rates and trends for the top 15 cancer sites [1], expressed as rates per 100,000 and age-adjusted.
The top five cancer killers (lung, colon, breast, pancreas and prostate) account for about 60% of all cancer deaths.
In the next blog, we'll cover funding for the common cancers.
1. [Seer Cancer Stat Fact Sheets. Table I-26 Age-adjusted U.S. death rates and trends for the top 15 cancer sites]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Though there has been little success in curing the advanced common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood. Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? In this blog, and in the next few blogs, we'll be exploring the properties of common tumors and how they differ from the properties of rare tumors.
First, lets look at common cancers and see how we've been approaching the problem of their cure.
The general approach to funding cancer research is the same now as it was in the early 1970s; attack cancer tumor by tumor, dividing available funds in rough proportion to the number of people who die from each tumor. Here are the top 15 cancer killers in the U.S. (Data from the National Cancer Institute's SEER, Surveillance Epidemiology and End Results Program).
List. U.S. death rates and trends for the top 15 cancer sites [1], expressed as rates per 100,000 and age-adjusted.
All Sites 189.8
Lung and Bronchus 54.1
Colon and Rectum 18.8
Breast 14.1
Pancreas 10.6
Prostate 10.1
Leukemia 7.4
Non-Hodgkin Lymphoma 7.3
Liver & IBD 5.0
Ovary 5.0
Esophagus 4.4
Brain and ONS 4.4
Urinary Bladder 4.3
Kidney and Renal Pelvis 4.2
Stomach 4.1
Myeloma 3.7
IBD = Intrahepatic bile duct
ONS = Other nervous system
The top five cancer killers (lung, colon, breast, pancreas and prostate) account for about 60% of all cancer deaths.
In the next blog, we'll cover funding for the common cancers.
1. [Seer Cancer Stat Fact Sheets. Table I-26 Age-adjusted U.S. death rates and trends for the top 15 cancer sites]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Tuesday, July 22, 2008
Neoplasms: 15
This is the fifteenth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
In an earlier blog, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress. The subsequent blogs explained these biases in some detail. This is the summarizing blog on the topic of survival data interpretation.
An advanced cancer is a cancer that has invaded extensively at its site of origin (i.e., into adjacent organs or into large vessels) or that has metastasized to other sites. Surgeons have been pretty good at curing cancers that are not advanced (i.e., cancers that can be completely resected at their primary sites of growth, before they have metastasized). For the most common cancers of adults (lung, breast, prostate, colon, pancreas, esophagus, liver, and so on), we have not had much luck curing the advanced cancers.
When oncologists discuss improved outcomes for the advanced stage common cancers, they are seldom referring a patient's chances of surviving the cancer. More often than not, they are discussing the length of time that a patient is expected to live following the diagnosis of the cancer. Though survival times are incrementally increasing, clinical trial data cannot support the conclusion that we are near to finding a cure for any of the advanced common cancers.
The only data we have that tells us anything about our progress in the war to conquer cancer is the age-adjusted cancer death rate. This data would indicate that the cumulative progress in preventing, diagnosing and treating cancer, over the past half century, has been small. To some people, these small improvements are encouraging. To others, it has been a sign that we are moving in the wrong direction.
Though there has been little success in curing the advanced common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood.
Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? Is it just a coincidence, or is there some fundamental principle involved?
In the next few blogs in this series on neoplasia, we will explore this question.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
In an earlier blog, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress. The subsequent blogs explained these biases in some detail. This is the summarizing blog on the topic of survival data interpretation.
An advanced cancer is a cancer that has invaded extensively at its site of origin (i.e., into adjacent organs or into large vessels) or that has metastasized to other sites. Surgeons have been pretty good at curing cancers that are not advanced (i.e., cancers that can be completely resected at their primary sites of growth, before they have metastasized). For the most common cancers of adults (lung, breast, prostate, colon, pancreas, esophagus, liver, and so on), we have not had much luck curing the advanced cancers.
When oncologists discuss improved outcomes for the advanced stage common cancers, they are seldom referring a patient's chances of surviving the cancer. More often than not, they are discussing the length of time that a patient is expected to live following the diagnosis of the cancer. Though survival times are incrementally increasing, clinical trial data cannot support the conclusion that we are near to finding a cure for any of the advanced common cancers.
The only data we have that tells us anything about our progress in the war to conquer cancer is the age-adjusted cancer death rate. This data would indicate that the cumulative progress in preventing, diagnosing and treating cancer, over the past half century, has been small. To some people, these small improvements are encouraging. To others, it has been a sign that we are moving in the wrong direction.
Though there has been little success in curing the advanced common cancers, there has been remarkable success in finding cures for some of the rare cancers, particularly several rare cancers of childhood.
Why is it possible to cure rare cancers? Why would rarity have anything to do with curability? Is it just a coincidence, or is there some fundamental principle involved?
In the next few blogs in this series on neoplasia, we will explore this question.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Monday, July 21, 2008
Neoplasms: 14
This is the fourteenth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's continue yesterday's comments on the observations of Dr. Ioannidis [1,2].
Dr. Ioannidis has published a list of conditions that reduce the likelihood that a research finding is valid [1]. Here is his list, with some small modifications.
-1. Small studies are less likely to produce true research findings than large studies.
-2. Small effects are less likely to be true than larger effects.
-3. Research findings are more likely to be true in confirmatory studies (such as phase III trials that confirm observations made in a phase II trial) than in hypothesis-generating studies.
-4. The greater the flexibility in design, definition, and measured outcome, the less likely that the research findings will be true.
-5. The greater the financial and other interests in a study, the less likely that the results will be true.
-6. The hotter the scientific field, the less likely that the results will be true.
What does all this mean? In the cancer field, should we not trust clinical trials? And if we can't trust clinical trials, how can we make any progress?
My opinion is that clinical trials are very important, but their conclusions should be considered tentative, not final.
Prospective randomized clinical trials are types of experiments and have all the vulnerabilities inherent in clinical studies. They can be poorly designed, misinterpreted, invalid for under-represented patient subpopulations, un-repeatable, and falsified. The best validation of predictive tests will come from continuous clinical correlations with patient outcomes in medical centers wherein many different types of patients (male, female, different nationalities, different ages, concurrent diseases, multiple medications) will be managed. In particular, survival data need to be validated by post-trial epidemiologic data. The data that are being interpreted must be made available to the public, and the conclusions should be scrutinized and debated.
1. [Ioannidis JP. Why most published research findings are false. PLoS Med 2:e124, 2005.]
2. [Ioannidis JP. Some main problems eroding the credibility and relevance of randomized trials. Bull NYU Hosp Jt Dis 66:135-139, 2008.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's continue yesterday's comments on the observations of Dr. Ioannidis [1,2].
Dr. Ioannidis has published a list of conditions that reduce the likelihood that a research finding is valid [1]. Here is his list, with some small modifications.
-1. Small studies are less likely to produce true research findings than large studies.
-2. Small effects are less likely to be true than larger effects.
-3. Research findings are more likely to be true in confirmatory studies (such as phase III trials that confirm observations made in a phase II trial) than in hypothesis-generating studies.
-4. The greater the flexibility in design, definition, and measured outcome, the less likely that the research findings will be true.
-5. The greater the financial and other interests in a study, the less likely that the results will be true.
-6. The hotter the scientific field, the less likely that the results will be true.
What does all this mean? In the cancer field, should we not trust clinical trials? And if we can't trust clinical trials, how can we make any progress?
My opinion is that clinical trials are very important, but their conclusions should be considered tentative, not final.
Prospective randomized clinical trials are types of experiments and have all the vulnerabilities inherent in clinical studies. They can be poorly designed, misinterpreted, invalid for under-represented patient subpopulations, un-repeatable, and falsified. The best validation of predictive tests will come from continuous clinical correlations with patient outcomes in medical centers wherein many different types of patients (male, female, different nationalities, different ages, concurrent diseases, multiple medications) will be managed. In particular, survival data need to be validated by post-trial epidemiologic data. The data that are being interpreted must be made available to the public, and the conclusions should be scrutinized and debated.
1. [Ioannidis JP. Why most published research findings are false. PLoS Med 2:e124, 2005.]
2. [Ioannidis JP. Some main problems eroding the credibility and relevance of randomized trials. Bull NYU Hosp Jt Dis 66:135-139, 2008.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Sunday, July 20, 2008
Neoplasms: 13
This is the thirteenth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Statistical Method Bias.
Strange as it may seem, a statistician can look at a set of data, apply different statistical methods to the data, and arrive at any of several different conclusions. Often, conclusions drawn by different methods, from different data sets are contradictory. In this case, articles with opposite conclusions appear in the medical literature, permitting scientists to selectively cite those papers that support their own agendas [1]. Depending on the method of choice, it is often possible to draw opposite conclusions from the same set of data.
John P.A. Ioannidis is the chair of the Clinical and Molecular Epidemiology Unit at the University of Ioannina School of Medicine and Biomedical Research Institute in Greece. In a provocative article entitled, "Why most published research findings are false," he points some common misinterpretations that pose as clinical facts [2]. These include: post hoc subgroup selection and analyses (i.e., cherry-picking a subgroup that qualifies for statistical significance); changing clinical group inclusion or exclusion criteria and disease definitions after the trial has concluded; selective or purposefully distorted reporting of results; data dredging (sifting through study data, searching for outlier groups); for multi-center studies, reporting the significant findings from some of the centers and ignoring negative results from other centers [2,3].
1. [Tatsioni A, Bonitsis NG, Ioannidis JP. Persistence of contradicted claims in the literature. JAMA 2007 Dec 5;298(21):2517-2526, 2007.]
2. [Ioannidis JP. Why most published research findings are false. PLoS Med 2:e124, 2005.]
3. [Ioannidis JP. Some main problems eroding the credibility and relevance of randomized trials. Bull NYU Hosp Jt Dis 66:135-139, 2008.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Statistical Method Bias.
Strange as it may seem, a statistician can look at a set of data, apply different statistical methods to the data, and arrive at any of several different conclusions. Often, conclusions drawn by different methods, from different data sets are contradictory. In this case, articles with opposite conclusions appear in the medical literature, permitting scientists to selectively cite those papers that support their own agendas [1]. Depending on the method of choice, it is often possible to draw opposite conclusions from the same set of data.
John P.A. Ioannidis is the chair of the Clinical and Molecular Epidemiology Unit at the University of Ioannina School of Medicine and Biomedical Research Institute in Greece. In a provocative article entitled, "Why most published research findings are false," he points some common misinterpretations that pose as clinical facts [2]. These include: post hoc subgroup selection and analyses (i.e., cherry-picking a subgroup that qualifies for statistical significance); changing clinical group inclusion or exclusion criteria and disease definitions after the trial has concluded; selective or purposefully distorted reporting of results; data dredging (sifting through study data, searching for outlier groups); for multi-center studies, reporting the significant findings from some of the centers and ignoring negative results from other centers [2,3].
1. [Tatsioni A, Bonitsis NG, Ioannidis JP. Persistence of contradicted claims in the literature. JAMA 2007 Dec 5;298(21):2517-2526, 2007.]
2. [Ioannidis JP. Why most published research findings are false. PLoS Med 2:e124, 2005.]
3. [Ioannidis JP. Some main problems eroding the credibility and relevance of randomized trials. Bull NYU Hosp Jt Dis 66:135-139, 2008.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Saturday, July 19, 2008
Neoplasms: 12
This is the twelfth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Medical Record and Re-abstraction Biases.
Trialists draw information related to the diagnosis, treatment, and outcome of patients by reviewing medical records. The quality of medical research often depends on the quality of medical records. When medical records are incomplete, incorrect, illegible, or otherwise uninterpretable, the results for an otherwise well-planned clinical trial can be disastrous.
Cause of death data comes from death certificates RfreaR. Death certificate data have many deficiencies [1,2]. The most common error occurs when a mode of death is listed as the cause of death. For example, cardiac arrest is not a cause of death, though it appears as the cause of death on many death certificates. There is not much value in a death certificate for a man who died with end-stage cancer when the listed cause of death is "cardia arrest." An international survey has shown very little consistency in the way that death data are collected [3]. Death certificates are completed without the benefit of an autopsy. At best, death certificates express a clinician's reasonable judgement at the time of a patient's death.
What do researchers do when they find that their medical records are inadequate. Often, they resort to re-abstraction, a time-consuming, expensive and occasionally futile undertaking. Re-abstraction often involves revisiting charts, visiting outpatient clinics and the private offices of medical doctors, re-interviewing patients and families, and a host of extraordinary efforts aimed at restoring credibility to clinical trial data. If a subset of patients has better maintained records than another subset, a bias can be introduced to the trial.
1. [Ashworth TG: Inadequacy of death certification: proposal for change. J Clin Pathol 44:265, 1991. Comment. A British perspective on the importance of the death certificate.]
2. [Kircher T, Anderson RE: Cause of death: proper completion of the death certificate. JAMA 258:349-352, 1987. Comment. Though every physician is expected to complete death certificates, surprisingly few physicians understand how to do the job. As a consequence, death certificates are notoriously inadequate records of the cause of death. The authors explain the differences between the underlying and immediate causes of death, between the mechanism and manner of death. They also describe how to complete the medical certification section of the death certificate.]
3. [Walter SD, Birnie SE: Mapping mortality and morbidity patterns: an international comparison. Intl J Epidemiology 20:678-689, 1991. Comment. This survey of 49 national and international health atlases has shown that there is virtually no consistency in the way that death data are presented.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Medical Record and Re-abstraction Biases.
Trialists draw information related to the diagnosis, treatment, and outcome of patients by reviewing medical records. The quality of medical research often depends on the quality of medical records. When medical records are incomplete, incorrect, illegible, or otherwise uninterpretable, the results for an otherwise well-planned clinical trial can be disastrous.
Cause of death data comes from death certificates RfreaR. Death certificate data have many deficiencies [1,2]. The most common error occurs when a mode of death is listed as the cause of death. For example, cardiac arrest is not a cause of death, though it appears as the cause of death on many death certificates. There is not much value in a death certificate for a man who died with end-stage cancer when the listed cause of death is "cardia arrest." An international survey has shown very little consistency in the way that death data are collected [3]. Death certificates are completed without the benefit of an autopsy. At best, death certificates express a clinician's reasonable judgement at the time of a patient's death.
What do researchers do when they find that their medical records are inadequate. Often, they resort to re-abstraction, a time-consuming, expensive and occasionally futile undertaking. Re-abstraction often involves revisiting charts, visiting outpatient clinics and the private offices of medical doctors, re-interviewing patients and families, and a host of extraordinary efforts aimed at restoring credibility to clinical trial data. If a subset of patients has better maintained records than another subset, a bias can be introduced to the trial.
1. [Ashworth TG: Inadequacy of death certification: proposal for change. J Clin Pathol 44:265, 1991. Comment. A British perspective on the importance of the death certificate.]
2. [Kircher T, Anderson RE: Cause of death: proper completion of the death certificate. JAMA 258:349-352, 1987. Comment. Though every physician is expected to complete death certificates, surprisingly few physicians understand how to do the job. As a consequence, death certificates are notoriously inadequate records of the cause of death. The authors explain the differences between the underlying and immediate causes of death, between the mechanism and manner of death. They also describe how to complete the medical certification section of the death certificate.]
3. [Walter SD, Birnie SE: Mapping mortality and morbidity patterns: an international comparison. Intl J Epidemiology 20:678-689, 1991. Comment. This survey of 49 national and international health atlases has shown that there is virtually no consistency in the way that death data are presented.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Friday, July 18, 2008
Neoplasms: 11
This is the eleventh blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Second Trial and Stage Treatment Biases.
After a therapeutic trial, clinicians can determine the type of patients who are most likely to benefit from an intervention. For example, a first trial of bone marrow transplantation for patients with metastatic carcinoma may indicate that patients over the age of 55 have very poor response to transplantation. Older individuals with transplants may be more prone to die from the interventional procedure than from their cancer.
On the second trial of the procedure, the clinicians will wisely exclude patients over some determined age (55 in this case). The second trial shows markedly improved survival, compared with the first trial, for those patients receiving bone marrow transplantation. The improvement can be achieved simply through better selection of subjects, and without improving the treatment protocol.
Stage treatment bias is closely related to second-trial bias. If you carefully select a stage of disease that is successfully treated by a particular treatment protocol, you can exaggerate the benefits of your treatment by ignoring disease stages for which your treatment is ineffective.
An example is the use of prostatectomy for prostate cancer, a procedure that is credited with a high cure rate. Prostatectomy is only performed on patients with tumor confined to the prostate. If the prostate cancer has metastasized to lymph nodes in the region of the prostate or to distant organs, prostatectomy is contra-indicated. Why is this? If the cancer has spread beyond the prostate, removing the prostate will not benefit the patient. An apt analogy is closing the barn doors after the horses have fled the farm.
Prostate cancer confined to the prostate is often indolent. By the age 80 to 90 years, 70% to 90% of men have prostate cancer confirmed at autopsy [1,2]. This indicates that prostate cancer is a very common disease that kills only a small proportion of affected individuals. Because prostatectomy is only performed on men whose prostate cancer is believed to be confined to the prostate, the cure rate is high. Restricting treatment to patients who have a stage of disease that is indolent, in most cases, virtually guarantees high survival rates.
1. [Guileyardo JM, Johnson WD, Welsh RA, Akazaki K, Correa P. Prevalence of latent prostate carcinoma in two U.S. populations. J Natl Cancer Inst 65:311-316, 1980.]
2. [Sakr WA, Grignon DJ, Haas GP, Heilbrun LK, Pontes JE, Crissman JD. Age and racial distribution of prostatic intraepithelial neoplasia. Eur Urol 30:138-144, 1996.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Science is not a collection of facts. Science is what facts teach us; what we can learn about our universe, and ourselves, by deductive thinking. From observations of the night sky, made without the aid of telescopes, we can deduce that the universe is expanding, that the universe is not infinitely old, and why black holes exist. Without resorting to experimentation or mathematical analysis, we can deduce that gravity is a curvature in space-time, that the particles that compose light have no mass, that there is a theoretical limit to the number of different elements in the universe, and that the earth is billions of years old. Likewise, simple observations on animals tell us much about the migration of continents, the evolutionary relationships among classes of animals, why the nuclei of cells contain our genetic material, why certain animals are long-lived, why the gestation period of humans is 9 months, and why some diseases are rare and other diseases are common. In “Armchair Science”, the reader is confronted with 129 scientific mysteries, in cosmology, particle physics, chemistry, biology, and medicine. Beginning with simple observations, step-by-step analyses guide the reader toward solutions that are sometimes startling, and always entertaining. “Armchair Science” is written for general readers who are curious about science, and who want to sharpen their deductive skills.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Second Trial and Stage Treatment Biases.
After a therapeutic trial, clinicians can determine the type of patients who are most likely to benefit from an intervention. For example, a first trial of bone marrow transplantation for patients with metastatic carcinoma may indicate that patients over the age of 55 have very poor response to transplantation. Older individuals with transplants may be more prone to die from the interventional procedure than from their cancer.
On the second trial of the procedure, the clinicians will wisely exclude patients over some determined age (55 in this case). The second trial shows markedly improved survival, compared with the first trial, for those patients receiving bone marrow transplantation. The improvement can be achieved simply through better selection of subjects, and without improving the treatment protocol.
Stage treatment bias is closely related to second-trial bias. If you carefully select a stage of disease that is successfully treated by a particular treatment protocol, you can exaggerate the benefits of your treatment by ignoring disease stages for which your treatment is ineffective.
An example is the use of prostatectomy for prostate cancer, a procedure that is credited with a high cure rate. Prostatectomy is only performed on patients with tumor confined to the prostate. If the prostate cancer has metastasized to lymph nodes in the region of the prostate or to distant organs, prostatectomy is contra-indicated. Why is this? If the cancer has spread beyond the prostate, removing the prostate will not benefit the patient. An apt analogy is closing the barn doors after the horses have fled the farm.
Prostate cancer confined to the prostate is often indolent. By the age 80 to 90 years, 70% to 90% of men have prostate cancer confirmed at autopsy [1,2]. This indicates that prostate cancer is a very common disease that kills only a small proportion of affected individuals. Because prostatectomy is only performed on men whose prostate cancer is believed to be confined to the prostate, the cure rate is high. Restricting treatment to patients who have a stage of disease that is indolent, in most cases, virtually guarantees high survival rates.
1. [Guileyardo JM, Johnson WD, Welsh RA, Akazaki K, Correa P. Prevalence of latent prostate carcinoma in two U.S. populations. J Natl Cancer Inst 65:311-316, 1980.]
2. [Sakr WA, Grignon DJ, Haas GP, Heilbrun LK, Pontes JE, Crissman JD. Age and racial distribution of prostatic intraepithelial neoplasia. Eur Urol 30:138-144, 1996.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Science is not a collection of facts. Science is what facts teach us; what we can learn about our universe, and ourselves, by deductive thinking. From observations of the night sky, made without the aid of telescopes, we can deduce that the universe is expanding, that the universe is not infinitely old, and why black holes exist. Without resorting to experimentation or mathematical analysis, we can deduce that gravity is a curvature in space-time, that the particles that compose light have no mass, that there is a theoretical limit to the number of different elements in the universe, and that the earth is billions of years old. Likewise, simple observations on animals tell us much about the migration of continents, the evolutionary relationships among classes of animals, why the nuclei of cells contain our genetic material, why certain animals are long-lived, why the gestation period of humans is 9 months, and why some diseases are rare and other diseases are common. In “Armchair Science”, the reader is confronted with 129 scientific mysteries, in cosmology, particle physics, chemistry, biology, and medicine. Beginning with simple observations, step-by-step analyses guide the reader toward solutions that are sometimes startling, and always entertaining. “Armchair Science” is written for general readers who are curious about science, and who want to sharpen their deductive skills.
Thursday, July 17, 2008
Neoplasms: 10
This is the tenth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Marketing and Money Bias.
Bevacizumab (developed and sold by Genentech as Avastin) is one of the most popular cancer drugs in the world and has been heralded as a wonder drug. It can easily cost $50,000 to $100,000 per year of use. One of the most exciting features of Avastin is that it can potentially treat any kind of cancer. Avastin is an antibody that works by attaching to VEGF (Vascular Endothelial Growth Factor), thus reducing the ability of tumors to vascularize and grow. In responsive cancers, studies indicate that it extends life by up to four months [1].
You may not be impressed by a drug that seems to extend life by up to four months in some responsive cancer. To the best of my knowledge, nobody has claimed that Avastin will cure advanced cancers. Why is Avastin popular? A well-marketed drug that promises hope for cancer patients can have enormous appeal to desperate patients, and their oncologists.
Are the results of clinical studies skewed in favor of the corporate sponsors of the trials? In a fascinating meta-analysis, Yank and coworkers wanted to know whether the results of clinical trials conducted with financial ties to a drug company, were biased towards favorable results [2]. They reviewed the literature on clinical trials for anti-hypertensive agents, and found that ties to a drug company did not bias the results. However, the found that financial ties to a drug company are associated with favorable conclusions. This suggests that regardless of the results of a trial, the conclusions published by the investigators were more likely to be favorable, if the trial were financed by a drug company. This should not be surprising. Two scientists can look at the same results and draw entirely different conclusions. It happens every day. How could an investigators, financed by a drug company, not be influenced by their benefactors when they interpret their results?
1. [Kolata G. Pollack A. Costly cancer drug offers hope, but also a dilemma. The New York Times, July 6, 2008.]
2. [Yank V, Rennie D, Bero LA. Financial ties and concordance between results and conclusions in meta-analyses: retrospective cohort study. BMJ 335:1202-1205, 2007.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at Marketing and Money Bias.
Bevacizumab (developed and sold by Genentech as Avastin) is one of the most popular cancer drugs in the world and has been heralded as a wonder drug. It can easily cost $50,000 to $100,000 per year of use. One of the most exciting features of Avastin is that it can potentially treat any kind of cancer. Avastin is an antibody that works by attaching to VEGF (Vascular Endothelial Growth Factor), thus reducing the ability of tumors to vascularize and grow. In responsive cancers, studies indicate that it extends life by up to four months [1].
You may not be impressed by a drug that seems to extend life by up to four months in some responsive cancer. To the best of my knowledge, nobody has claimed that Avastin will cure advanced cancers. Why is Avastin popular? A well-marketed drug that promises hope for cancer patients can have enormous appeal to desperate patients, and their oncologists.
Are the results of clinical studies skewed in favor of the corporate sponsors of the trials? In a fascinating meta-analysis, Yank and coworkers wanted to know whether the results of clinical trials conducted with financial ties to a drug company, were biased towards favorable results [2]. They reviewed the literature on clinical trials for anti-hypertensive agents, and found that ties to a drug company did not bias the results. However, the found that financial ties to a drug company are associated with favorable conclusions. This suggests that regardless of the results of a trial, the conclusions published by the investigators were more likely to be favorable, if the trial were financed by a drug company. This should not be surprising. Two scientists can look at the same results and draw entirely different conclusions. It happens every day. How could an investigators, financed by a drug company, not be influenced by their benefactors when they interpret their results?
1. [Kolata G. Pollack A. Costly cancer drug offers hope, but also a dilemma. The New York Times, July 6, 2008.]
2. [Yank V, Rennie D, Bero LA. Financial ties and concordance between results and conclusions in meta-analyses: retrospective cohort study. BMJ 335:1202-1205, 2007.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Wednesday, July 16, 2008
Neoplasms: 9
This is the ninth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at population and demographic biases.
For many different reasons, some populations accrue more easily into clinical trials than other groups. Most notorious are children and pregnant women. Many clinical trials have no children and no pregnant women. Under such conditions, drug effectiveness and safety cannot extend to these groups. Third party payers may refuse to cover the costs of drugs for pregnant women and children due to lack of any trial evidence indicating that the drug is safe and effective in these groups.
Demographic bias is a variant of population bias. In most clinical trials in the U.S., patients are assigned broad demographic groups, and the patients are often allowed to assign themselves into groups based on the group to which they have the closest social identity. Treatment response rates that may differ from group to group and for unaccounted subgroups within a single group (e.g., Japanese patients may have a different response from Chinese patients).
In clinical trials, groupings are usually based on stage of disease and age. Seldom do clinical trials stratify patients by income. Nonetheless, socioeconomic status greatly influences cancer survival [1]. Groups that contain many economically disadvantaged patients are likely to have a shorter survival than similar groups where the members are financially well-off.
Population bias effects every population that is not included in study populations. A good examples comes from our interpretation of PSA (prostate specific antigen) values in men. PSA is the most important screening test for prostatic cancer. PSA levels less than 4 indicate low risk of prosate cancer. PSA levels greater than 4 indicate high-risk patients and prompt a diagnostic study. The clinically accepted ranges of PSA levels were adapted from data collected from a community-based Minnesota population consisting entirely of white men [2]. In discussing the importance of age-specific ranges, the Mayo Clinic group acknowledged that African African-Americans and Asians were omitted from their study [3].
Does the screening criteria, developed for normal PSA ranges in a white population, fit the normal ranges for African-Americans and Asians? In a review of PSA levels among white men and black men, conducted by Sawyer and coworkers, studying 30,000 PSA values collected for about 3,000 subjects, it was shown that African-Americans have a higher range of PSA levels than do white Americans [4]. A separate study of PSA levels in Asian men found a lower range of PSA levels in this population [5].
1. [Gorey KM, Holowary EJ, Fehringer G, Laukkanen E, Moskowitz A, Webster DJ, Richter NL. An international comparison of cancer survival: Toronto, Ontario, and Detroit, Michigan, metropolitan areas. Am J Public Health 87:1156-1163, 1997.]
2. [Oesterling JE, Jacobsen ST, Chute CG, Guess HA, Girman CJ, Panser LA, Lieber MM. Serum prostate-specific antigen in a community-based population of healthy men. JAMA 270:860-864, 1993.]
3. [Oesterling JD, Jacobsen SJ, Cooner WH. The use of age-specific reference ranges for serum prostate specific antigen in men 60 years old or older. J Urol 153:1160-1163, 1995.]
4. [Sawyer R, Berman JJ, Borkowski A, Moore GW. Elevated prostate-specific antigen levels in black men and white men. Modern Pathology 9:1029-1032, 1996.]
5. [Oesterling JE, Kumamoto Y, Tsukamoto T, Girman CJ, Guess HA, Masumori N, Jacobsen SJ, Lieber MM. Serum prostate specific antigen in a community-based population of healthy Japanese men: lower values than for similarly aged white men. Br J Urol 75:347-352, 1995.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at population and demographic biases.
For many different reasons, some populations accrue more easily into clinical trials than other groups. Most notorious are children and pregnant women. Many clinical trials have no children and no pregnant women. Under such conditions, drug effectiveness and safety cannot extend to these groups. Third party payers may refuse to cover the costs of drugs for pregnant women and children due to lack of any trial evidence indicating that the drug is safe and effective in these groups.
Demographic bias is a variant of population bias. In most clinical trials in the U.S., patients are assigned broad demographic groups, and the patients are often allowed to assign themselves into groups based on the group to which they have the closest social identity. Treatment response rates that may differ from group to group and for unaccounted subgroups within a single group (e.g., Japanese patients may have a different response from Chinese patients).
In clinical trials, groupings are usually based on stage of disease and age. Seldom do clinical trials stratify patients by income. Nonetheless, socioeconomic status greatly influences cancer survival [1]. Groups that contain many economically disadvantaged patients are likely to have a shorter survival than similar groups where the members are financially well-off.
Population bias effects every population that is not included in study populations. A good examples comes from our interpretation of PSA (prostate specific antigen) values in men. PSA is the most important screening test for prostatic cancer. PSA levels less than 4 indicate low risk of prosate cancer. PSA levels greater than 4 indicate high-risk patients and prompt a diagnostic study. The clinically accepted ranges of PSA levels were adapted from data collected from a community-based Minnesota population consisting entirely of white men [2]. In discussing the importance of age-specific ranges, the Mayo Clinic group acknowledged that African African-Americans and Asians were omitted from their study [3].
Does the screening criteria, developed for normal PSA ranges in a white population, fit the normal ranges for African-Americans and Asians? In a review of PSA levels among white men and black men, conducted by Sawyer and coworkers, studying 30,000 PSA values collected for about 3,000 subjects, it was shown that African-Americans have a higher range of PSA levels than do white Americans [4]. A separate study of PSA levels in Asian men found a lower range of PSA levels in this population [5].
1. [Gorey KM, Holowary EJ, Fehringer G, Laukkanen E, Moskowitz A, Webster DJ, Richter NL. An international comparison of cancer survival: Toronto, Ontario, and Detroit, Michigan, metropolitan areas. Am J Public Health 87:1156-1163, 1997.]
2. [Oesterling JE, Jacobsen ST, Chute CG, Guess HA, Girman CJ, Panser LA, Lieber MM. Serum prostate-specific antigen in a community-based population of healthy men. JAMA 270:860-864, 1993.]
3. [Oesterling JD, Jacobsen SJ, Cooner WH. The use of age-specific reference ranges for serum prostate specific antigen in men 60 years old or older. J Urol 153:1160-1163, 1995.]
4. [Sawyer R, Berman JJ, Borkowski A, Moore GW. Elevated prostate-specific antigen levels in black men and white men. Modern Pathology 9:1029-1032, 1996.]
5. [Oesterling JE, Kumamoto Y, Tsukamoto T, Girman CJ, Guess HA, Masumori N, Jacobsen SJ, Lieber MM. Serum prostate specific antigen in a community-based population of healthy Japanese men: lower values than for similarly aged white men. Br J Urol 75:347-352, 1995.]
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Tuesday, July 15, 2008
Neoplasms: 8
This is the eighth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at confounder bias.
Confounders are unanticipated or ignored factors that alter an outcome measurement. It is impossible to design clinical trials that account for confounder influences because most confounders are unanticipated. Confounders are the statistical byproducts of the "Law of unanticipated consequences," which simply asserts that there will always be unanticipated consequences.
Statins are a widely used drugs that reduce the blood levels of cholesterol and various other blood lipids. To the best of my knowledge, nobody expected that the use of statins would have any effect on the incidence or mortality of cancer.
In a recent study involving nearly a half-million male patients conducted between 1998 and 2004, statin use exceeding six months was linked to a significant lung cancer risk reduction of 55%. Participants who took a statin longer than four years had a 77% reduction in lung cancer risk [Khurana]. Let us imagine that the report is accurate and that we can eliminate deaths from lung cancer by 77% just by prescribing statins to everyone at risk for lung cancer. In the United States, about 90,000 men and 70,000 women will die of lung cancer this year [Jemal]. If this were reduced by 77%, we would prevent the cancer deaths of about 123,000 people.
Given these surprising findings, who knows what the effects of statins may be on a cancer treatment trial? If there were unanticipated influences of statin use among some clinical trial participants, who take statins for reasons unrelated to the trial design, then this would be just one example of a confounder bias.
Khurana V, Bejjanki HR, Caldito G, Owens MW. Statins reduce the risk of lung cancer in humans: a large case-control study of US veterans. Chest 131:1282-1288, 2007.]
Jemal A, Murray T, Ward E, et al. Cancer statistics, 2005. CA Cancer J Clin 55:10-30, 2005.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at confounder bias.
Confounders are unanticipated or ignored factors that alter an outcome measurement. It is impossible to design clinical trials that account for confounder influences because most confounders are unanticipated. Confounders are the statistical byproducts of the "Law of unanticipated consequences," which simply asserts that there will always be unanticipated consequences.
Statins are a widely used drugs that reduce the blood levels of cholesterol and various other blood lipids. To the best of my knowledge, nobody expected that the use of statins would have any effect on the incidence or mortality of cancer.
In a recent study involving nearly a half-million male patients conducted between 1998 and 2004, statin use exceeding six months was linked to a significant lung cancer risk reduction of 55%. Participants who took a statin longer than four years had a 77% reduction in lung cancer risk [Khurana]. Let us imagine that the report is accurate and that we can eliminate deaths from lung cancer by 77% just by prescribing statins to everyone at risk for lung cancer. In the United States, about 90,000 men and 70,000 women will die of lung cancer this year [Jemal]. If this were reduced by 77%, we would prevent the cancer deaths of about 123,000 people.
Given these surprising findings, who knows what the effects of statins may be on a cancer treatment trial? If there were unanticipated influences of statin use among some clinical trial participants, who take statins for reasons unrelated to the trial design, then this would be just one example of a confounder bias.
Khurana V, Bejjanki HR, Caldito G, Owens MW. Statins reduce the risk of lung cancer in humans: a large case-control study of US veterans. Chest 131:1282-1288, 2007.]
Jemal A, Murray T, Ward E, et al. Cancer statistics, 2005. CA Cancer J Clin 55:10-30, 2005.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Monday, July 14, 2008
Neoplasms: 7
This is the seventh blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at diagnosis bias.
Diagnosis bias occurs when a group of people who have a disease or lesion, that is different from the disease or lesion carried by the previously studied group, is used in a repeat clinical trial. This distorts the survival outcome after treatment.
For example, if patients were accrued into a study based on erroneous diagnoses rendered by an incompetent pathologist, a group of patients with breast cancer might become diluted with patients who have benign breast disease. It is possible to generate an inflated survival rate, compared with earlier studies that did not include erroneous diagnoses, by including healthy patients in the treatment group.
Most modern clinical trials impose strict pathology reviews in an effort to reduce the effect of diagnosis bias.
More commonly, treatment groups are tainted when a new subset of patients is added to the original group, through enhanced detection and diagnosis of a biologically distinctive variant of disease. For example, there has been a many-fold increase in the detection and diagnosis of ductal carcinoma in situ (DCIS) of breast, over the past few decades, thanks to mass mammographic screenings. DCIS has a very good prognosis. Only a small percentage of patients with DCIS progress to invasive ductal carcinoma.
As the proportion of breast cancer patients with diagnosed DCIS increases, the survival of the group that includes DCIS cases also increases. By increasing the number of patients with a good prognosis breast cancer, an improved survival can be achieved, without an improvement in therapeutic protocol.
The remedy for diagnosis bias is to subdivide your study groups carefully. This remedy has its own problems. If the first study did not carefully separate patients by their prognostic subtypes, how can you compare prior results with future results (where the subtypes were separated). Also, if you break groups by their tumor subtypes, will each group contain a sufficient number of patients to produce statistically useful results?
We'll continue discussing the biases in survival data in the next blog.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at diagnosis bias.
Diagnosis bias occurs when a group of people who have a disease or lesion, that is different from the disease or lesion carried by the previously studied group, is used in a repeat clinical trial. This distorts the survival outcome after treatment.
For example, if patients were accrued into a study based on erroneous diagnoses rendered by an incompetent pathologist, a group of patients with breast cancer might become diluted with patients who have benign breast disease. It is possible to generate an inflated survival rate, compared with earlier studies that did not include erroneous diagnoses, by including healthy patients in the treatment group.
Most modern clinical trials impose strict pathology reviews in an effort to reduce the effect of diagnosis bias.
More commonly, treatment groups are tainted when a new subset of patients is added to the original group, through enhanced detection and diagnosis of a biologically distinctive variant of disease. For example, there has been a many-fold increase in the detection and diagnosis of ductal carcinoma in situ (DCIS) of breast, over the past few decades, thanks to mass mammographic screenings. DCIS has a very good prognosis. Only a small percentage of patients with DCIS progress to invasive ductal carcinoma.
As the proportion of breast cancer patients with diagnosed DCIS increases, the survival of the group that includes DCIS cases also increases. By increasing the number of patients with a good prognosis breast cancer, an improved survival can be achieved, without an improvement in therapeutic protocol.
The remedy for diagnosis bias is to subdivide your study groups carefully. This remedy has its own problems. If the first study did not carefully separate patients by their prognostic subtypes, how can you compare prior results with future results (where the subtypes were separated). Also, if you break groups by their tumor subtypes, will each group contain a sufficient number of patients to produce statistically useful results?
We'll continue discussing the biases in survival data in the next blog.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Sunday, July 13, 2008
Neoplasms: 6
This is the sixth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at lead time bias.
Suppose there were a cancer, cancer Y, that is uniformly deadly. Once it is diagnosed, the average survival, after the best available treatment, is three years. Nobody who has this cancer lives beyond five years.
Dr. Detecto is a pathologist who has invented a very sensitive method for detecting cancer Y at a very early stage. Dr. Detecto can detect cancer Y a full four years earlier than any previous method of detection. Unfortunately, there is no effective treatment for cancer Y, even when it is detected early. All patients with cancer Y will die. Because cancer Y patients are now detected four years earlier, the natural course of disease results in an expected death 7 years (3 years plus the 4 years lead time) later. When we study 5-year survival after diagnosis, we find that the five year survival is now 90%.
The newspaper headline reads, "New, improved detection technique for cancer Y improves 5-year survival from 0% to 90%."
Of course, detecting the cancer four years earlier only increased the time between diagnosis and death. It did not extend, by even a single minute, the age at death of patients with cancer Y.
Has Dr. Detecto made a useless discovery, and is the survival data fraudulent? No. Tumors are best treated when they are detected early. In the case of cancer Y, there was no immediate benefit for early detection. Nonetheless, the set of early cancers provided cancer researchers with a group of tumors that might have an improved response to alternate cancer therapies. This would require clinical trials. The survival data from the early-diagnosed group can be very misleading. It is the responsibility of trialists and journalists to interpret survival data cautiously.
We'll continue discussing the biases in survival data in the next blog.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Several days ago, I listed over a dozen biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Today, let's look at lead time bias.
Suppose there were a cancer, cancer Y, that is uniformly deadly. Once it is diagnosed, the average survival, after the best available treatment, is three years. Nobody who has this cancer lives beyond five years.
Dr. Detecto is a pathologist who has invented a very sensitive method for detecting cancer Y at a very early stage. Dr. Detecto can detect cancer Y a full four years earlier than any previous method of detection. Unfortunately, there is no effective treatment for cancer Y, even when it is detected early. All patients with cancer Y will die. Because cancer Y patients are now detected four years earlier, the natural course of disease results in an expected death 7 years (3 years plus the 4 years lead time) later. When we study 5-year survival after diagnosis, we find that the five year survival is now 90%.
The newspaper headline reads, "New, improved detection technique for cancer Y improves 5-year survival from 0% to 90%."
Of course, detecting the cancer four years earlier only increased the time between diagnosis and death. It did not extend, by even a single minute, the age at death of patients with cancer Y.
Has Dr. Detecto made a useless discovery, and is the survival data fraudulent? No. Tumors are best treated when they are detected early. In the case of cancer Y, there was no immediate benefit for early detection. Nonetheless, the set of early cancers provided cancer researchers with a group of tumors that might have an improved response to alternate cancer therapies. This would require clinical trials. The survival data from the early-diagnosed group can be very misleading. It is the responsibility of trialists and journalists to interpret survival data cautiously.
We'll continue discussing the biases in survival data in the next blog.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Saturday, July 12, 2008
Neoplasms: 5
This is the fifth blog in a series of blogs on neoplasia.
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Yesterday, I listed 15 biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Of the sources of bias listed yesterday, stage assignment bias is the most subtle and the most difficult to understand. It is also the most important bias introduced by the most modern cancer trials, which are often aimed at developing stage-specific treatments for groups of patients with common cancers.
Suppose every patient with X type of cancer is staged into one of two groups (I and II). The stage I group has no evidence of distant metastases at the time of diagnosis and has a 40% chance of 5-year survival under the standard treatment protocol. Patients are put into the Stage II group if they have distant metastases at the time of diagnosis. Their chance of having a 5-year survival under the standard treatment is 2%.
Professor Rads, at the University of Goodcare, has recently developed a very sensitive imaging device that can detect small metastases that would be undetectable by less sophisticated devices. In the next clinical trial for treatment of cancer X, Professor Rads tests each trial candidate with his new device. He finds that about half of the patients who would otherwise be assigned to Stage I (no metastases) are found to have metastases with his sensitive machine. With this information, these erstwhile Stage I patients are re-assigned into Stage II.
A clinical trial is conducted with the standard treatment. The Stage I group is now much smaller than the Stage II group. When the trial is complete, we find that the 5-year survival for the Stage I group is now 80% (up from 40%). The 5-year survival for the Stage II group is now 2% (the same as before).
The newspaper headline following the trial is: "New imaging discovery yields 100% improvement in survival for Stage I Cancer X"
Actually, the clinical trial, as described, yielded no improved survival for any patients. All it accomplished was to correctly re-assign some of the Stage I patients into the low-survival Stage II group. The apparent improvement in survival in Stage I cancer patients was simply the result of more accurate staging of patients in the low-risk category of disease.
What does this mean? Was the clinical trial a fraud? Did it accomplish nothing at all? No, accurate staging of patients with cancer is an absolutely crucial step in the development of new, effective anti-cancer regimens. It is impossible to assess the effects of a new chemotherapeutic agent on a group with heterogeneous disease. The beneficial effects of a new drug on Stage I cancer patients might be lost in a study where the Stage I group is mixed with patients with Stage II cancers (that might not be responsive to the regimen).
By accurately staging patients with cancer, trialists can test and develop drugs that are most likely to benefit patients in a particular stage of disease. In fact, that is the current approach to cancer trials: subdividing cancer patients into well-defined clinical subsets and fine-tuning new anti-cancer regimens to maximize their effectiveness for each subset of disease.
So we see, in this instance, that cancer survival data is important, but it can be misleading.
We'll continue discussing the biases in survival data in the next blog.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In the past few blogs, I've been trying to explain the disconnect between cancer survival data and cancer death rate data. The cancer survival data seems to indicate that we're making enormous improvements in cancer treatment. The cancer death rate indicates that Americans are dying from cancer at about the same rate as they had been a half-century ago.
Yesterday, I listed 15 biases in cancer survival data that contribute to an overly optimistic sense of medical progress.
In this and the next few blogs, I thought I'd review some of these biases. The purpose of this exercise is to explain that the interpretation of survival data is enormously complex and that survival data is probably not the best way to gauge progress in the field of cancer research.
Of the sources of bias listed yesterday, stage assignment bias is the most subtle and the most difficult to understand. It is also the most important bias introduced by the most modern cancer trials, which are often aimed at developing stage-specific treatments for groups of patients with common cancers.
Suppose every patient with X type of cancer is staged into one of two groups (I and II). The stage I group has no evidence of distant metastases at the time of diagnosis and has a 40% chance of 5-year survival under the standard treatment protocol. Patients are put into the Stage II group if they have distant metastases at the time of diagnosis. Their chance of having a 5-year survival under the standard treatment is 2%.
Professor Rads, at the University of Goodcare, has recently developed a very sensitive imaging device that can detect small metastases that would be undetectable by less sophisticated devices. In the next clinical trial for treatment of cancer X, Professor Rads tests each trial candidate with his new device. He finds that about half of the patients who would otherwise be assigned to Stage I (no metastases) are found to have metastases with his sensitive machine. With this information, these erstwhile Stage I patients are re-assigned into Stage II.
A clinical trial is conducted with the standard treatment. The Stage I group is now much smaller than the Stage II group. When the trial is complete, we find that the 5-year survival for the Stage I group is now 80% (up from 40%). The 5-year survival for the Stage II group is now 2% (the same as before).
The newspaper headline following the trial is: "New imaging discovery yields 100% improvement in survival for Stage I Cancer X"
Actually, the clinical trial, as described, yielded no improved survival for any patients. All it accomplished was to correctly re-assign some of the Stage I patients into the low-survival Stage II group. The apparent improvement in survival in Stage I cancer patients was simply the result of more accurate staging of patients in the low-risk category of disease.
What does this mean? Was the clinical trial a fraud? Did it accomplish nothing at all? No, accurate staging of patients with cancer is an absolutely crucial step in the development of new, effective anti-cancer regimens. It is impossible to assess the effects of a new chemotherapeutic agent on a group with heterogeneous disease. The beneficial effects of a new drug on Stage I cancer patients might be lost in a study where the Stage I group is mixed with patients with Stage II cancers (that might not be responsive to the regimen).
By accurately staging patients with cancer, trialists can test and develop drugs that are most likely to benefit patients in a particular stage of disease. In fact, that is the current approach to cancer trials: subdividing cancer patients into well-defined clinical subsets and fine-tuning new anti-cancer regimens to maximize their effectiveness for each subset of disease.
So we see, in this instance, that cancer survival data is important, but it can be misleading.
We'll continue discussing the biases in survival data in the next blog.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Friday, July 11, 2008
Neoplasms: 4
This is the fourth blog in a series of blogs on neoplasia.
In prior blogs in this series, we reviewed NCI data on changes in U.S. cancer death rate since 1950. There has been very little improvement in the U.S. cancer death rate in more than a half century. In my opinion, death rate is a good measure of progress in the war against cancer because it reflects cumulative changes from cancer causes, cancer preventions, and cancer treatments.
Though the cancer death rate has not decreased much in the past half century, there are many recent claims that cancer survival has greatly improved.
Cancer survival is determined by finding the proportion of people who are alive at various intervals after receiving a diagnosis of cancer. The claims are that today, a person with a diagnosis of cancer would be expected to survive the cancer much longer than someone receiving the same diagnosis at some earlier time.
Here is an excerpt from the MedicineNet site:
http://www.medicinenet.com/script/main/art.asp?articlekey=157
The title of the report is, "Better and Longer Survival for Cancer Patients"
The lead-off text is, "Statistics (released in 1997) show that cancer patients are living longer and even "beating" the disease. Information released at an AMA sponsored conference for science writers, showed that the death rate from the dreaded disease has decreased by 3 percent in the last few years. In the 1940's only one patient in four survived on the average. By the 1960's, that figure was up to one in three, and now has reached 50% survival."
When conducting a clinical trial for a new cancer drug, it is essential to show that the drug offers some kind of improvement over the standard treatment. Determining differences in cancer survival between the standard treatment group and the experimental treatment group is very important. In well-designed trials, these comparisons answer very specific questions about specific subsets of people with specific clinical stages of disease.
It is difficult, or impossible, to generalize much about the survivability of all stages of cancer, in all cancer patients, by looking at survival data in clinical trials.
Here is a partial listing of the biases in survival data:
1. Stage assignment bias (diagnostic or screening tools shifting the proportion of people at different disease stages)
2. Lead-time bias (extending time-after diagnosis without changing date of death)
3. Population bias (exclusion of important subpopulations)
4. Statistical method bias (different methods in different studies of same treatment)
5. Demographic bias (different demographics in the time interval of the study)
6. Measurement bias (inability to accurately measure clinical or biological study parameters)
7. Record bias (medical records can be incomplete flawed, including death status)
8. Re-abstraction bias (gaps in records often require re-abstraction from other sources, and there may be biases in the way that the additional information is collected for certain subpopulations of patients).
9. Comorbidity bias (survial may be determined by processes other than the cancer under study).
10. Overdiagnosis bias (when the patient has a benign condition that is erroneously diagnosed as cancer)
11. Second trial bias (also called patient selection bias. Second trials of the same drug or procedure commonly produce better survival results than the first trial, because the clinicians become more adept at selecting patients who will benefit from the treatment).
12. Money bias (if there's money at stake, even minimal benefit can be spun into major advances; even 4 month extensions in life can be promoted as miracles)
13. Stage treatment bias (finding an improvement that is effective for a small subset of people with a cancer can be falsely interpreted as a measure that enhances survival for everyone).
14. Apples-Oranges bias (sometimes you can't determine improvements in survival because the objectives of different studies may not be comparable)
15. Confounders (sometimes factors may improve health and survivability without being part of a treatment; for example, statins may increase survival by reducing the risk of heart disease in cancer patients, without having any direct effect on the patient's cancer).
For all these reasons, you need to be very careful when discussing survival trends in cancer patients. I'll discuss some of these specific points (about survival studies) in future blogs.
My opinion is that if there were major gains in cancer survival, they would show up in the U.S. cancer death rate. As discussed in prior blogs in this series, there has been only incremental (at best) improvement in the U.S. cancer death rate in the past half-century.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In June, 2014, my book, entitled Rare Diseases and Orphan Drugs: Keys to Understanding and Treating the Common Diseases was published by Elsevier. The book builds the argument that our best chance of curing the common diseases will come from studying and curing the rare diseases.
I urge you to read more about my book. There's a generous preview of the book at the Google Books site. If you like the book, please request your librarian to purchase a copy of this book for your library or reading room.
In prior blogs in this series, we reviewed NCI data on changes in U.S. cancer death rate since 1950. There has been very little improvement in the U.S. cancer death rate in more than a half century. In my opinion, death rate is a good measure of progress in the war against cancer because it reflects cumulative changes from cancer causes, cancer preventions, and cancer treatments.
Though the cancer death rate has not decreased much in the past half century, there are many recent claims that cancer survival has greatly improved.
Cancer survival is determined by finding the proportion of people who are alive at various intervals after receiving a diagnosis of cancer. The claims are that today, a person with a diagnosis of cancer would be expected to survive the cancer much longer than someone receiving the same diagnosis at some earlier time.
Here is an excerpt from the MedicineNet site:
http://www.medicinenet.com/script/main/art.asp?articlekey=157
The title of the report is, "Better and Longer Survival for Cancer Patients"
The lead-off text is, "Statistics (released in 1997) show that cancer patients are living longer and even "beating" the disease. Information released at an AMA sponsored conference for science writers, showed that the death rate from the dreaded disease has decreased by 3 percent in the last few years. In the 1940's only one patient in four survived on the average. By the 1960's, that figure was up to one in three, and now has reached 50% survival."
When conducting a clinical trial for a new cancer drug, it is essential to show that the drug offers some kind of improvement over the standard treatment. Determining differences in cancer survival between the standard treatment group and the experimental treatment group is very important. In well-designed trials, these comparisons answer very specific questions about specific subsets of people with specific clinical stages of disease.
It is difficult, or impossible, to generalize much about the survivability of all stages of cancer, in all cancer patients, by looking at survival data in clinical trials.
Here is a partial listing of the biases in survival data:
1. Stage assignment bias (diagnostic or screening tools shifting the proportion of people at different disease stages)
2. Lead-time bias (extending time-after diagnosis without changing date of death)
3. Population bias (exclusion of important subpopulations)
4. Statistical method bias (different methods in different studies of same treatment)
5. Demographic bias (different demographics in the time interval of the study)
6. Measurement bias (inability to accurately measure clinical or biological study parameters)
7. Record bias (medical records can be incomplete flawed, including death status)
8. Re-abstraction bias (gaps in records often require re-abstraction from other sources, and there may be biases in the way that the additional information is collected for certain subpopulations of patients).
9. Comorbidity bias (survial may be determined by processes other than the cancer under study).
10. Overdiagnosis bias (when the patient has a benign condition that is erroneously diagnosed as cancer)
11. Second trial bias (also called patient selection bias. Second trials of the same drug or procedure commonly produce better survival results than the first trial, because the clinicians become more adept at selecting patients who will benefit from the treatment).
12. Money bias (if there's money at stake, even minimal benefit can be spun into major advances; even 4 month extensions in life can be promoted as miracles)
13. Stage treatment bias (finding an improvement that is effective for a small subset of people with a cancer can be falsely interpreted as a measure that enhances survival for everyone).
14. Apples-Oranges bias (sometimes you can't determine improvements in survival because the objectives of different studies may not be comparable)
15. Confounders (sometimes factors may improve health and survivability without being part of a treatment; for example, statins may increase survival by reducing the risk of heart disease in cancer patients, without having any direct effect on the patient's cancer).
For all these reasons, you need to be very careful when discussing survival trends in cancer patients. I'll discuss some of these specific points (about survival studies) in future blogs.
My opinion is that if there were major gains in cancer survival, they would show up in the U.S. cancer death rate. As discussed in prior blogs in this series, there has been only incremental (at best) improvement in the U.S. cancer death rate in the past half-century.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
In June, 2014, my book, entitled Rare Diseases and Orphan Drugs: Keys to Understanding and Treating the Common Diseases was published by Elsevier. The book builds the argument that our best chance of curing the common diseases will come from studying and curing the rare diseases.
I urge you to read more about my book. There's a generous preview of the book at the Google Books site. If you like the book, please request your librarian to purchase a copy of this book for your library or reading room.
Thursday, July 10, 2008
Neoplasms: 3
This is the third blog in a series of blogs on neoplasia.
As shown in yesterday's blog, the death rate from cancer in the U.S. is about the same today as it was in 1950 and 1978.
The cancer death rate is the number of deaths from cancer (per 100,000 age-adjusted population). The cancer death rate is very different from the cancer survival rate. The cancer survival rate is determined by the number of people who have cancer and who do not die from the cancer (in some pre-determined observation period).
The cancer death rate is affected by factors that have nothing to do with cancer treatment. For example, if the incidence of a major cancer were to drop (as in the case of lung cancer in certain subpopulations), this might reduce the cancer death rate, even if it were not accompanied by improved methods of lung cancer treatment. Likewise, if the population started to die off from some cause other than cancer (e.g., a major influenza outbreat), the cancer death rate might drop. We'll discuss some of the limitations of the cancer death rate in a future blog. Still, as a rough indicator, the cancer death rate serves an important purpose. The U.S. cancer death rate has barely budged in over 50 years, and this tells us that we are not winning the war against cancer.
In the past year, I've noticed that cancer journalists like to point out that the cancer death rate has been steadily dropping since the early 1990s. This is true, but the importance of this trend is small.
Let's look at the year-by-year cancer death rate from NCI's Surveillance Epidemiology and End Results project, available at:
http://seer.cancer.gov/csr/1975_2005/results_merged/topic_annualrates.pdf
The cancer death rate has been dropping since 1991, but only after a small rise in the cancer death rate from 1975 to 1991. Basically, we have about the same cancer death rate today as we had in 1975. Nobody can say, with any certainty, why we had that bump in the U.S. cancer death rate between 1975 and 2005. But cancer researchers did not take credit for the rise in cancer prior to 1991. Likewise, cancer researchers cannot take credit for the drop in cancer after 1991.
All we can say with any confidence is that our progress against cancer has not yielded any major changes in the U.S. cancer death rate.
What is the significance of the cancer survival rate, and what can say about our ability to successfully treat people who develop cancer? That will be the subject of the next few blogs.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
As shown in yesterday's blog, the death rate from cancer in the U.S. is about the same today as it was in 1950 and 1978.
The cancer death rate is the number of deaths from cancer (per 100,000 age-adjusted population). The cancer death rate is very different from the cancer survival rate. The cancer survival rate is determined by the number of people who have cancer and who do not die from the cancer (in some pre-determined observation period).
The cancer death rate is affected by factors that have nothing to do with cancer treatment. For example, if the incidence of a major cancer were to drop (as in the case of lung cancer in certain subpopulations), this might reduce the cancer death rate, even if it were not accompanied by improved methods of lung cancer treatment. Likewise, if the population started to die off from some cause other than cancer (e.g., a major influenza outbreat), the cancer death rate might drop. We'll discuss some of the limitations of the cancer death rate in a future blog. Still, as a rough indicator, the cancer death rate serves an important purpose. The U.S. cancer death rate has barely budged in over 50 years, and this tells us that we are not winning the war against cancer.
In the past year, I've noticed that cancer journalists like to point out that the cancer death rate has been steadily dropping since the early 1990s. This is true, but the importance of this trend is small.
Let's look at the year-by-year cancer death rate from NCI's Surveillance Epidemiology and End Results project, available at:
http://seer.cancer.gov/csr/1975_2005/results_merged/topic_annualrates.pdf
Year Adjusted cancer death rate per 100,000
1975 199.1
1976 202.3
1977 203.0
1978 204.4
1979 204.5
1980 207.0
1981 206.4
1982 208.3
1983 209.2
1984 210.9
1985 211.3
1986 211.8
1987 211.9
1988 212.6
1989 214.3
1990 214.9
1991 215.1
1992 213.5
1993 213.4
1994 211.7
1995 209.9
1996 207.0
1997 203.6
1998 200.8
1999 200.7
2000 198.7
2001 195.9
2002 193.7
2003 190.0
2004 185.8
2005 184.0
The cancer death rate has been dropping since 1991, but only after a small rise in the cancer death rate from 1975 to 1991. Basically, we have about the same cancer death rate today as we had in 1975. Nobody can say, with any certainty, why we had that bump in the U.S. cancer death rate between 1975 and 2005. But cancer researchers did not take credit for the rise in cancer prior to 1991. Likewise, cancer researchers cannot take credit for the drop in cancer after 1991.
All we can say with any confidence is that our progress against cancer has not yielded any major changes in the U.S. cancer death rate.
What is the significance of the cancer survival rate, and what can say about our ability to successfully treat people who develop cancer? That will be the subject of the next few blogs.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Wednesday, July 9, 2008
Neoplasms: 2
This is the second blog in a series of blogs on neoplasia.
Let us admit that we are not winning the war on cancer.
Here are the facts, published by the U.S. National Cancer Institute's SEER (Surveillance, Epidemiology and End Results) project, and available at:
http://seer.cancer.gov/csr/1975_2005/
results_merged/topic_historical_mort_trends.pdf
The numbers are deaths per 100,000 population. These numbers are age-adjusted. We'll talks about age adjustment of incidence rates in a future blog, but for the moment, it suffices to note that age-adjustments compensates for incidence biases that would apply if there were large age-shifts in the populations (1950, 1978, 2005).
For all ages, the death rate from cancer rose from 195.4 in 1950 to 204.4 in 1975 and then dropped a little to 184.0 in 2005. Much has been made of the drop in death rate between 1975 and 2005. Actually, the drop takes us to a number that is very similar to the death rate back in 1950.
If you look at the different age groups, the biggest drops occur in the pediatric ages. This confirms what we already know, that we have made remarkable gains in treating childhood cancer. However, childhood cancer is rare. The large drop in the incidence of childhood cancer is just a drop in the bucket of the large number of cancer deaths occurring in adults.
With few exceptions, adults are dying from the same types of cancers today that were killing adults in 1950, and at about the same rates.
How can this be true when we are told by experts that cancer survival is improving?
Here is an excerpt from the MedicineNet site:
http://www.medicinenet.com/script/main/art.asp?articlekey=157
The title of the report is, "Better and Longer Survival for Cancer Patients"
The lead-off text is, "Statistics (released in 1997) show that cancer patients are living longer and even "beating" the disease. Information released at an AMA sponsored conference for science writers, showed that the death rate from the dreaded disease has decreased by 3 percent in the last few years. In the 1940's only one patient in four survived on the average. By the 1960's, that figure was up to one in three, and now has reached 50% survival."
When we look at death rates, it seems that our risk of dying from cancer is about the same today as it was in 1950. Contrariwise, when we talk about cancer survival, it looks as though we've made remarkable progress.
What's going on?
In the next few blogs, I'll discuss the kinds of lessons we learn from looking at cancer death rates and the kinds of information we get when we study cancer survival.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Let us admit that we are not winning the war on cancer.
Here are the facts, published by the U.S. National Cancer Institute's SEER (Surveillance, Epidemiology and End Results) project, and available at:
http://seer.cancer.gov/csr/1975_2005/
results_merged/topic_historical_mort_trends.pdf
Table I-2
56-YEAR TRENDS IN U.S. CANCER DEATH RATES
All Races, Males and Females
Age Group 1950 1978 2005
0-4 11.1 4.6 2.2
5-14 6.7 4.1 2.5
15-24 8.6 6.1 4.1
25-34 20.4 14.2 9.1
35-44 63.6 50.7 32.8
45-54 174.2 179.6 118.3
55-64 391.3 428.9 329.7
65-74 710.0 803.4 748.8
75-84 1167.2 1204.1 1265.1
85+ 1450.7 1535.3 1643.7
All Ages 195.4 204.4 184.0
The numbers are deaths per 100,000 population. These numbers are age-adjusted. We'll talks about age adjustment of incidence rates in a future blog, but for the moment, it suffices to note that age-adjustments compensates for incidence biases that would apply if there were large age-shifts in the populations (1950, 1978, 2005).
For all ages, the death rate from cancer rose from 195.4 in 1950 to 204.4 in 1975 and then dropped a little to 184.0 in 2005. Much has been made of the drop in death rate between 1975 and 2005. Actually, the drop takes us to a number that is very similar to the death rate back in 1950.
If you look at the different age groups, the biggest drops occur in the pediatric ages. This confirms what we already know, that we have made remarkable gains in treating childhood cancer. However, childhood cancer is rare. The large drop in the incidence of childhood cancer is just a drop in the bucket of the large number of cancer deaths occurring in adults.
With few exceptions, adults are dying from the same types of cancers today that were killing adults in 1950, and at about the same rates.
How can this be true when we are told by experts that cancer survival is improving?
Here is an excerpt from the MedicineNet site:
http://www.medicinenet.com/script/main/art.asp?articlekey=157
The title of the report is, "Better and Longer Survival for Cancer Patients"
The lead-off text is, "Statistics (released in 1997) show that cancer patients are living longer and even "beating" the disease. Information released at an AMA sponsored conference for science writers, showed that the death rate from the dreaded disease has decreased by 3 percent in the last few years. In the 1940's only one patient in four survived on the average. By the 1960's, that figure was up to one in three, and now has reached 50% survival."
When we look at death rates, it seems that our risk of dying from cancer is about the same today as it was in 1950. Contrariwise, when we talk about cancer survival, it looks as though we've made remarkable progress.
What's going on?
In the next few blogs, I'll discuss the kinds of lessons we learn from looking at cancer death rates and the kinds of information we get when we study cancer survival.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Tuesday, July 8, 2008
Neoplasms: 1
I have written a book, Neoplasms: Principles of Development and Diversity, which will be published October 1, 2008. Amazon and Barnes and Noble are already taking pre-publication orders for the book.
Neoplasms explains what causes cancer, how cancer develops, and why we see the types of neoplasms (tumors) that occur in man and other animals (tumor speciation). Once we understand the principles of tumor development and speciation, we can build a biological classification of neoplasms. The last section of the book shows how we can use a classification of neoplasms to develop class-based methods to eliminate cancers and precancers (through prevention, diagnosis and treatment). To regular readers of this blog, these are all famililar themes.
When I was writing the book, there were many subjects that I had to cut out from the final text. These consisted mostly of complex topics that could not be included due to length limitations (the book is 464 pages, as it is).
Over the next few months, I will be writing a long series of blogs containing material that was excluded from the book or that was included in a shortened form. This blog should be of interest to cancer specialists and to biomedical informaticians who want to see how tools of the post-information age (ontologies, public databases, mathematical models, collective data analysis) can be be used to conquer diseases.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Neoplasms explains what causes cancer, how cancer develops, and why we see the types of neoplasms (tumors) that occur in man and other animals (tumor speciation). Once we understand the principles of tumor development and speciation, we can build a biological classification of neoplasms. The last section of the book shows how we can use a classification of neoplasms to develop class-based methods to eliminate cancers and precancers (through prevention, diagnosis and treatment). To regular readers of this blog, these are all famililar themes.
When I was writing the book, there were many subjects that I had to cut out from the final text. These consisted mostly of complex topics that could not be included due to length limitations (the book is 464 pages, as it is).
Over the next few months, I will be writing a long series of blogs containing material that was excluded from the book or that was included in a shortened form. This blog should be of interest to cancer specialists and to biomedical informaticians who want to see how tools of the post-information age (ontologies, public databases, mathematical models, collective data analysis) can be be used to conquer diseases.
-Copyright (C) 2008 Jules J. Berman
key words: cancer, tumor, tumour, carcinogen, neoplasia, neoplastic development, classification, biomedical informatics, tumor development, precancer, benign tumor, ontology, classification, developmental lineage classification and taxonomy of neoplasms
Saturday, July 5, 2008
Bomedicine in the Post-Information Age: 7 of 7
This is the seventh of 7 blogs on biomedicine in the post-information age.
In the post-information age, there is universal access to information, computational power, and the world-wide communications infrastructure.
The first point I've been trying to make in this series of 7 blogs is that the "work" of the post-information age is to derive meaning from our ubiquitous information. The second point I've been trying to make is that individuals, rather than institutions, are in the best position to make the most rapid and the most startling advances in this new age.
Throughout this series of blogs, I've mentioned data annotation, without explaining what I meant. In its simplest form, data annotation is adding metadata (data descriptors) to the data in a document. The purpose of adding metadata is to make the data specific. So a date can be an item on a calendar, or a social event, or a type of fruit. Metadata allows you to specify your intent.
If a date is a fruit, then it is a type of organism:
ID : 42345
PARENT ID : 4719
RANK : species
GC ID : 1
MGC ID : 1
SCIENTIFIC NAME : Phoenix dactylifera
GENBANK COMMON NAME : date palm
SYNONYM : Phoenix dactylifera L.
HIERARCHY
Phoenix dactylifera
Phoenix
Phoeniceae
Coryphoideae
Arecaceae
Arecales
commelinids
Liliopsida
Magnoliophyta
Spermatophyta
Euphyllophyta
Tracheophyta
Embryophyta
Streptophytina
Streptophyta
Viridiplantae
Eukaryota
cellular organisms
The date grows on a date tree (Phoenix dactylifera) and inherits the properties of its ancestors. The organism ancestry (phylogeny) of the data was obtained at my web page, by entering Phoenix dactylifera in the query box.
http://www.julesberman.info/post.htm
By using metadata that is specified in a classification or an ontology, we can use annotated data to draw inferences that are beyond the intent of the original document. By merging annotated documents (a product of the information age), and applying post-information age data analysis tools, we can achieve a great deal.
The new age starts with data specification.
- Copyright (C) 2008 Jules J. Berman
key words: semantics, semantic web, RDF, biomedical informatics
Science is not a collection of facts. Science is what facts teach us; what we can learn about our universe, and ourselves, by deductive thinking. From observations of the night sky, made without the aid of telescopes, we can deduce that the universe is expanding, that the universe is not infinitely old, and why black holes exist. Without resorting to experimentation or mathematical analysis, we can deduce that gravity is a curvature in space-time, that the particles that compose light have no mass, that there is a theoretical limit to the number of different elements in the universe, and that the earth is billions of years old. Likewise, simple observations on animals tell us much about the migration of continents, the evolutionary relationships among classes of animals, why the nuclei of cells contain our genetic material, why certain animals are long-lived, why the gestation period of humans is 9 months, and why some diseases are rare and other diseases are common. In “Armchair Science”, the reader is confronted with 129 scientific mysteries, in cosmology, particle physics, chemistry, biology, and medicine. Beginning with simple observations, step-by-step analyses guide the reader toward solutions that are sometimes startling, and always entertaining. “Armchair Science” is written for general readers who are curious about science, and who want to sharpen their deductive skills.
In the post-information age, there is universal access to information, computational power, and the world-wide communications infrastructure.
The first point I've been trying to make in this series of 7 blogs is that the "work" of the post-information age is to derive meaning from our ubiquitous information. The second point I've been trying to make is that individuals, rather than institutions, are in the best position to make the most rapid and the most startling advances in this new age.
Throughout this series of blogs, I've mentioned data annotation, without explaining what I meant. In its simplest form, data annotation is adding metadata (data descriptors) to the data in a document. The purpose of adding metadata is to make the data specific. So a date can be an item on a calendar, or a social event, or a type of fruit. Metadata allows you to specify your intent.
If a date is a fruit, then it is a type of organism:
ID : 42345
PARENT ID : 4719
RANK : species
GC ID : 1
MGC ID : 1
SCIENTIFIC NAME : Phoenix dactylifera
GENBANK COMMON NAME : date palm
SYNONYM : Phoenix dactylifera L.
HIERARCHY
Phoenix dactylifera
Phoenix
Phoeniceae
Coryphoideae
Arecaceae
Arecales
commelinids
Liliopsida
Magnoliophyta
Spermatophyta
Euphyllophyta
Tracheophyta
Embryophyta
Streptophytina
Streptophyta
Viridiplantae
Eukaryota
cellular organisms
The date grows on a date tree (Phoenix dactylifera) and inherits the properties of its ancestors. The organism ancestry (phylogeny) of the data was obtained at my web page, by entering Phoenix dactylifera in the query box.
http://www.julesberman.info/post.htm
By using metadata that is specified in a classification or an ontology, we can use annotated data to draw inferences that are beyond the intent of the original document. By merging annotated documents (a product of the information age), and applying post-information age data analysis tools, we can achieve a great deal.
The new age starts with data specification.
- Copyright (C) 2008 Jules J. Berman
key words: semantics, semantic web, RDF, biomedical informatics
Science is not a collection of facts. Science is what facts teach us; what we can learn about our universe, and ourselves, by deductive thinking. From observations of the night sky, made without the aid of telescopes, we can deduce that the universe is expanding, that the universe is not infinitely old, and why black holes exist. Without resorting to experimentation or mathematical analysis, we can deduce that gravity is a curvature in space-time, that the particles that compose light have no mass, that there is a theoretical limit to the number of different elements in the universe, and that the earth is billions of years old. Likewise, simple observations on animals tell us much about the migration of continents, the evolutionary relationships among classes of animals, why the nuclei of cells contain our genetic material, why certain animals are long-lived, why the gestation period of humans is 9 months, and why some diseases are rare and other diseases are common. In “Armchair Science”, the reader is confronted with 129 scientific mysteries, in cosmology, particle physics, chemistry, biology, and medicine. Beginning with simple observations, step-by-step analyses guide the reader toward solutions that are sometimes startling, and always entertaining. “Armchair Science” is written for general readers who are curious about science, and who want to sharpen their deductive skills.
Thursday, July 3, 2008
Biomedical Informatics "Search Inside"
Readers of this blog know that I published Biomedical Informatics (Copyright 2007) with Jones & Bartlett Publishers. This book has never been scanned by Google Books (for sample chapters), and was never given a "Search Inside" by Amazon. A "Search Inside" is a web site that features excerpts from the book, including the full table of contents.
Amazon has, at long last, produced a "Search Inside," for Biomedical Informatics. I hope that readers of this blog will visit the Amazon site.
- Jules Berman
Amazon has, at long last, produced a "Search Inside," for Biomedical Informatics. I hope that readers of this blog will visit the Amazon site.
- Jules Berman
Subscribe to:
Posts (Atom)