Monday, November 30, 2009
Slowly but inevitably we are moving from "Dr. knows Best" to "everybody is a little bit doctor", to next generation diagnostics, better software tools and databases, personal devices, wireless health tools, and new healthcare models.
The first milestone on this road is known as Health 2.0:
Monday, November 23, 2009
Recent mammogram debate shows why reform will fail - most women don't like to get mammograms later in life and less frequently. Of course, this furious reaction is fueled by those who would face payment cuts if the new guidelines are implemented - radiologists, for example.
How many medical tests do you need to sleep sound at night knowing that you are healthy enough? Billions, trillions, googols?
How many out of a few thousand existing medical tests are redundant or irrelevant to individual consumers?
Mamograms and colonoscopies are perfect examples of overused diagnostic procedures. (see also the consumerreports article on overused tests and treatments).
Colonoscopy accounts for up to 75% of costs of an IBS patient, while the probability of it yielding meaningful results is less than 3%.
Citing Dr. Rex: Low-risk patients might undergo too many colonoscopies; high-risk patients, too few. When given a range of hypothetical findings on initial colonoscopy, most physicians would recommend repeat colonoscopy earlier than is indicated by the U.S. Multi-Society Task Force guidelines.
If 2000 women are screened regularly for 10 years, one will benefit from the screening, as she will avoid dying from breast cancer.10 healthy women, however, will have either a part of their breast or the whole breast removed. Even statistics deny that early screening for breast cancer saves lives.
Early and frequent screenings often lead to false alarms and unneeded biopsies, without substantially improving women's odds of survival. About 90 percent of abnormal mammogram findings are benign. A 2009 study in the British Medical Journal estimated that roughly one in three breast cancers detected by mammograms would never have caused harm. Earlier studies were suggesting this too, raising (along with harmful consequences of false positive results)
Perhaps the biggest problem with performing too many screening tests in healthy people is a phenomenon called over-diagnosis: Screening can detect slow-growing, harmless cancers that would never have killed you. But even if a mammogram "involves a tiny dose of radiation", a bigger dose used in radiotherapy is harmful when given to healthy people. Breast cancer radiotherapy regimens can increase mortality from heart disease and lung cancer 10-20 years afterwards.
How often does this happen with breast cancer?
Dr. Kevin Pho, an internist in Nashua, N.H., thinks the rebellion indicates a larger problem with ObamaCare. "The fact that [the administration] is distancing itself from what I consider to be very robust guidelines portends a very poor future for comparative effectiveness," he says. "If they back down now, what's going to happen when a comparative effectiveness body says there's no difference between angioplasty and medical management of heart disease?"
The American College of Obstetrics and Gynecology (ACOG) has just revised their guidelines for Pap smears under some pressure. This resulted from an Annals of Internal Medicine article which documented that only 16.4% of gynecologists followed the College’s prior guidelines. Most did more screenings than indicated, the worst record of the specialties tested. But the ACOG still recommends that nearly all women obtain regular screening at intervals of 1-3 years.
Cervical cancer is a rare disease in the US: just over 11,000 cases are predicted in 2009. There will be nearly as many cases of testicular cancer, 8,400. In comparison both breast and prostate cancer are just under 200,000. Most women have been led to believe that cervical cancer is rampant and they need yearly screening to prevent it. Testicular cancer however, is rarely mentioned. Most physicians don’t even bother to recommend that young men self-examine.Here is a different example. Man sues over “botched” testicular surgery. Doctors later discovered that the tumor was not malignant and did not need to be removed. Was it possible to offer a better testing, say biopsy? Urologists can tell you that testes should never be biopsied prior to removal if cancer is suspected as it significantly increases the risk of tumor dissemination. Studies show that those who had scrotal incisions for biopsy have a higher local recurrence rate as well as a higher relapse rate. If there is a solid growth in the testes, there is 95-97% probability of the growth being malignant. Sounds very reasonable, in this case doctors are sued for proper care, but was another cost - the cost of inconvenience, perhaps even mental trauma - ever taken into consideration?
Medicine is as much art as science – there are many cases where there are no “right” answers. Health care providers can prescribe as many procedures as they want and charge whatever they like. It's time for this to be changed.
Personal risk of cervical cancer, for example, can be easily estimated by every woman as it depends on whether she had multiple sexual partners, prior negative Paps, long term mutually monogamous relationship. HIV (that has a five times greater incidence than cervical cancer) tests are not administered to people without risks, right? Well, not quite right, but that's another story.
Dr. Joel Sherman (Medical Privacy, A Patient Oriented Discussion) has seen many women who are angry that the facts on cervical cancer have been hidden from them. They are pushed into getting Paps, but never told the pros and cons of screening. Never mentioned are the high incidence of abnormalities that resolve spontaneously, like negative biopsies and colposcopies.
And last, but not least - another reason of why healthcare is so costly - the insurer's overestimates, in a way their overdiagnosis of our health problems.
If the law says insurers have to treat every person the same, without taking into account whether they’re sick or healthy, young or old, a rational insurer will do some rational things. For example, it will assume disproportionate numbers of people who buy a policy from them will be sick and old. Of course, when they do this, the product becomes expensive, and young, healthy people start to wonder if they should even buy it in the first place. After all, they don’t really need insurance, right? Coupled with the overall rise in the cost of health care, insurers now push through new rounds of price increases, which, in turn, create more uninsured people. It is a very nasty cycle.
Aurametrix is developing decision support systems to help you evaluate personal health risks, decide on preventative measures, estimate cost/benefits of performing diagnostic tests. Better tools for a healthier world.
Wednesday, November 18, 2009
Here are my brief meeting minutes from the Cofounders Wanted November Meetup, organized by @alain94040
(see also FairSoftware's blog on this event and Aurametrix' analysis of startups presenting at Silicon Valley New Technologies meetup)
- Interesting idea of hotels for everyone and by everyone was presented by russ.hearl at staysherpa.com The startup is looking for a CTO-to-be PHP programmer. http://www.sherpatravelexchange.com/ & http://twitter.com/sherpatravelx
- Greg Gentschev is looking for a technical cofounder for a Business Search & Workflows project: http://twitter.com/gentschev
- Justin at binarynomads.com is looking for a technical cofounder too: http://binarynomads.com/
- Peter Kazanjy is building a site to outperform LinkedIn and Yelp, his LI prfile is at http://www.linkedin.com/in/kazanjy see also http://getunvarnished.com/
- Ira Chayut is building a Fun Site with Sound and Voices: http://www.linkedin.com/in/irachayut
- Sujit Kirpekar (kirpekar at gmail.com) has a great Autolicious idea, follow him on twitter: http://twitter.com/discoganya
- Kristen at Hazard.com presented her Timesarrow project: http://www.hazardbio.com/, see also http://twitter.com/timesarrow. She is a yoga buff and is looking for a business cofounder.
- Georgi Dagnall at geogad hopes people will travel more, with their personal mobile tour guide. See http://www.geogad.com/
- Junk Cloud brings the power of auctions to online classified ads: http://www.junkcloud.com/ and http://twitter.com/JunkCloud. Contact Ryan at juncloud.com for more information
- Al Brown, Deep Therapeutics, is looking for mechanical, electronics, robotics engineers to help implement the prototype.
- http://fairsoftware.net/- legal framework for early-stage projects, finding cofounders
- http://www.chagora.com/ - microdonations
- http://revenzy.com/ - social auctions site
- As to Aurametrix, it's all hush hush now. The public site is for general informational purposes only. The company may be looking for a cofounder at a later stage.
Thursday, November 5, 2009
Microbiota is specific to every individual, and varies systematically across body habitats and time, as well as geographical location, preventing or causing a disease after exposures to infectious agents.
Some human skin locations harbor even more diverse bacterial communities than the gut that we were thoughtfully nourishing with probiotics.
New analysis published in Science Express adds more information to the earlier results (from May 2009, for example), showing how diverse the microbiota is and how easy it is to re-colonize the skin.
We mapped some of the findings as shown in the Figure (on the right; the figure on the left maps bacteria in GI tract, from Dr. Richard Lord’s presentation at the 2008 Functional Medicine Symposium in Carlsbad, CA). Moist sites are shown with blue arrows, such as inside the nose, the armpits, the navel, dry areas are shown with green arrows, such as the forearm and oily sites are shown with yellow arrows: inside the ear, between the eyebrows, forehead, the back of the scalp.
Sites of most bacterial diversity were : The index finger, back of knee, forearm, palm and sole of foot.The forehead displayed the least diversity (with bacterial populations strongly preffering this site and not letting other bacteria to co-habit the space), but there were individual differences between different people. The mouth cavity showed the least variation in diversity both within individuals and between people. Studies of other microbes such as viruses and bacteriophages show low diversity in the airways as well, even though the human respiratory tract is constantly exposed to a wide variety of microbes and environmental agents. There is a difference between diseased and non-diseased individuals though - in Cystic Fibrosis (CF). for example, viromes are enriched in aromatic amino acid metabolism. Note that this disease causes a distinct acidic breath - the more severe the condition is in an individual, the more acidic his breath becomes. The microbes were especially sensitive to amino-acid starvation indicating that therapeutic measures may be more effective if used to change the respiratory environment, as opposed to shifting the taxonomic composition of resident microbiota.
Altered breath resulting from changed micrflora is a known phenomenon and it can be detected not only by complex mass spec machines, but also by devices used in QA testing of foods (e.g. Cyranose pick up the scent of penzane, isoprene acetone, and benzene in the breath of lung cancer patients) and car air quality sensors to study human "fermentome". (See also ongoing clinical trials on chemicals in human breath for diagnostics of diseases).
Altered bacterial populations could, indeed, be studied by metabonomic profiling. At present, however, the most accurate analysis, was performed based on microbial DNA or 16S RNA.
The study subjects were sampled four times each over a three-month period, typically after showering an hour or two earlier. Microbial DNA was then isolated directly from swabs used for sampling each body site. To recover bacteria from the skin surface, it was enough to swab it once by a wet cotton swab in 30s.
Wednesday, November 4, 2009
Workflow, procedures and organization of diagnostic laboratories have changed little since the end of the 19th century. Technology improved quality and safety, lead to higher throughput and allowed private electronic access to patients' lab test results - at least partially, but other than that not much has changed.
The introduction of first generation transcriptome technologies in the mid 1990s (Schena et al., 1995; DeRisi et al., 1996) has led to a phenomenal ability to simultaneously measure thousands of genes, create molecular profiles of cancers, all other diseases and conditions.
Proteomics - coined a couple of years later (James, 1997) - was accepted as an even more promising technique for effective diagnostics of diseases. Figure on the left, however, demonstrates that it too faces many challenges from discovery of biomarkers to their verification and approval. After more than a decade (Oliver et al., 1998), metabolomics has been accepted as - at least - an equally promising technique, but it still lagging behin genomics and proteomics.
All these techniques will have significant impact on the business model of diagnostics. Multiplexing (measuring multiple biomarkers at once) is obviously much more cost-effective. Diagnostics industry is historically very resistant to disruptive technological change, but potential cost advantages should outweigh this, leading to novel business models in health management.
A change will also come from the growing near patient testing (NPT) sector. NPT is already finding a role in wellness monitoring. Existing self tests - such as cholesterol kits - may not be very accurate, but with the advent of inexpensive multiplexing assays this will be overcome. Even when blood testing is done by trained professionals in a lab, there can be significant variability in test results. Same applies to blood pressure measurements - you may need to do three measurements per day for five days in order to get a decent baseline. This only justifies the need to have inexpensive tests that can be done more often.
But lets go back to metabolomics and its potential to provide noninvasive inexpensive diagnostics. Are there any clinical trials attempting to translate it into clinical practice?
Here are our favorite ones:
NCT00757952: Diagnosing ovarian cancer in exhaled breath. (Pine Street Foundation & University of Maine)
NCT00898209: Diagnosing Lung cancer in exhaled breath. (Vanderbilt-Ingram Cancer Center)
NCT00898209: Exhaled breath analyzed for lung cancer. (Vanderbilt-Ingram Cancer Center)
NCT00639067: Breath test for early detection of lung cancer (Menssana Research)
NCT00873366: Breath tests to access effectiveness of breast cancer treatment (Mayo Clinic and National Cancer Institute (NCI))
NCT00330603: Metabolomic breath analysis to predict treatment for chronic cough (University of Virginia)
NCT00632307: Breath analysis to diagnose COPD; lung cancer; airway infection; interstitial lung disease, sleep apnea; pulmonary disorders with pleural infusions; sarcoidosis (Lung Clinic Hemer, Germany)
NCT00294489: Breath analysis to diagnose Hepatitis C (Hadassah Medical Organization, Jerusalem, Israel)
Schena M, Shalon D, Davis RW, Brown PO 1995 Quantitative monitoring of gene expression patterns with complementary DNA microarray. Science 270 : 467 –470
DeRisi J, Penland L, Brown PO, Bittner ML, Meltzer PS, Ray M, Chen Y, Su YA 1996 Use of a cDNA microarray to analyze gene expression patterns in human cancer. Nat Genet 14 : 457 –460[CrossRef][Medline]
James P 1997 Protein identification in the post-genome era: the rapid rise of proteomics.". Quarterly reviews of biophysics 30 (4): 279–331. doi:10.1017/S0033583597003399. PMID 9634650.
Oliver SG, Winson MK, Kell DB, Baganz F. 1998. Systematic functional analysis of the yeast genome. Trends in Biotechnology 16: 373-378.
Aurametrix is conducting research to develop next-generation diagnostics to help you evaluate your personal health risks and benefits. Better tools for a healthier world.
Sunday, November 1, 2009
Image by juhansonin via Flickr
ACM Silicon Valley Data Mining Camp on November 1, 2009 has attracted more than 200 people with different backgrounds and interests. It was held at Hacker’s Dojo, sponsored by REvolution Computing, KXEN (Knowledge Extraction Engines), and LinkedIn (See notes on this event by @Andraz of Zemanta, Ken's open source tools, and relevant #dmcamp twits) .
Biomedical/Healthcare data mining topic was suggested by Junling Hu and Irene Gabashvili and supported by A.J. Chen, Greg Makowski, Sukanta Ganguly, and 40 other participants of the Data Mining Camp. Below is a brief transcript of the discussion.
The session started from introductions, here are some of them:
- Irene, with background in biophysics, medical informatics and CS, pursuing a personal health management venture, interested in data mining to advance personalized medicine;
- Lawrence, with background in physics and software engineering and interest in health IT. He is the organizer of Google Wave meetup (you may know about Google Health Wave);
- Hua, formerly with Kaiser, interested in medical scheduling and web development;
- Liana, interested in Natural Language Processing for biomedical knowledge mining;
- Maura, interested in Health IT, medical engineering and security;
- Magnus, developing Medical Databases;
- Kevin, interested in medical startups;
- Watson, with background in genomics and machine learning;
- Peters, working on medical devices and software embedded systems;
- Steve, formerly of Applied Biosystems;
- Roy of Codexis, focusing on data mining and pattern recognition in multivariate time series
- Jima, with background in medical informatics;
- Karsleep, interested in biomedical data mining;
- Deena, scientific analyst interested in how data mining technologies could be applied to healthcare;
- Junling Hu, scientist at Bosch, working on a device and software collecting and analyzing patients' information, based on daily questionnaires and other collected data.
Junling started the session from mentioning a recently published paper on computer technologies for healthcare determining strategic directions in the area. Irene also suggested to check the mHealth Summit focusing on mobile technologies to improve research data collection, healthcare delivery, and health outcomes.
Junling described the project she was working on - inexpensive device collecting data and sending it to a "coaching" nurse that monitors stay-at-home patients. Next step is to mine the data automatically, thus reducing the load on healthcare professionals without sacrificing patients' well-being. Junling also mentioned some of the challenges such as compliance of participants who are typically not eager to fill out the 20-question surveys. This is especially bad for obesity studies.
The data mining challenges mentioned during the session were:
(1) Missing Data
We are not talking about sparse data (discussed in one of the previous sessions on data mining with R), but actually missing data. Data is sparse if only a small fraction of the attributes are non-null - like the number of items we typically buy in a grocery store is much less than the number of products they offer. Data is missing if the values were never entered or the member combination is not meaningful (for example, obstetrics/gynecolgy values not meaningful for men) . One of the suggestions from the experts in the audience was to utilize "multiple imputation". Other suggestions included "once-a-week" questioning instead of daily surveys. Irene mentioned the 7D-PAR (Seven-Day Physical Activity Recall) , one of standardized questionnaires developed in the 80s (1,2) and other established methods.
Questions and comments from the audience:
- Data mining methods utilized for Chronic Disease Assesment and Elderly monitoring. Junling talked about unsupervised classification algorithms and two supervised learning methods she found to be most useful for her work - SVM and logistic regression. Both were equally good in predicting hospitalization events
- Indicators of Goodness of Model Predictions. Suggested events were hospitalizations, mortality... It was noted that good indicators are yet to be found.
This was another health data mining challenge emphasized during the discussion. All standard methods can be applied such as accuracy, precision, recall, true positives, false positives and especially combinations of the last 2 measures. Junling mentioned breast cancer classifier developed by Siemens and other algorithms predicting emergency situations with 90% accuracy. Irene noted that one of the problems of digital mammography and other cancer predictors is a high rate of false positives. From 30 to 40% of cancers are overdiagnosed (3), thus increasing healthcare costs. This has to be changed.
Several people in the audience emphasized that existing methods are averaging the population. Medicine needs to be truly personalized, we need better methods and more data.
(3) Large Number of Input Features
One of the main problems of health data mining is coping up with large number of input features. Obviously, a 20-question test is not sufficient. Should it rely on thousand questions or trillion inputs? And how to select a subset of relevant features to build robust learning models? Junling's preffered approaches are logistic regression and singular value decomposition. She would add features one by one and check if the overall accuracy for predictions remains good.
Questions from the audience included:
- A 3-5 year Vision for Health Data Mining: what do we expect to achieve?
Participants expressed an optimistic outlook
- Ray: Are most input variables discrete or continuous? The answer was: mixed
The good thing about pattern recognition is that the more patterns you have, the better it performs. Google translator is a good proof of this assertion (although this translator needs even more patterns to do a decent job).
Biomedical data sets such as CT, MRI, PET scans and other image data, gene expression, genetic variation are very large scale in nature. The challenge for data miners is to integrate and extract information from data of such scale.
Questions from the audience:
- What are the other large-scale studies trying ot mine patterns in health data, outside of US?
Studies in China and Taiwan using similar devices and models; also in Europe
Adding to this interesting discussion that was unfortunately interrupted because of the lack of time, I'd like to mention a few other challenges facing biomedical data mining.
- We should not underestimate the complexity of relationships between causative and effect variables in human health. Simplistic approaches are deemed to fail. Over-fitting could be a problem too
- Integration between heterogeneous data sources and types,and putting content in context (semantic integration) remains a challenge.
- Privacy Concerns associated with the Sharing of Individual Health Information.
- Blair S. How to assess exercise training habit and physical fitness. In: Behavioral Health, edited by Matarazzo JD. New York: Wiley, 1984, p. 424-447.
- Rauramaa R., Tuomainen P., Väisänen S., and Rankinen T. Physical activity and health- related fitness in middle-aged men. Med Sci Sports Exerc 27: 707-712, 1995.
- Gøtzsche, P.C., Jørgensen, K.J., Mæhlen, J. and Zahl, P.-H. Estimation of lead time and overdiagnosis in breast cancer screening. British Journal of Cancer (2009) 100, 219–219.