Sunday, November 23, 2014

Genomic Screening


Share

My greatest interest is in the correlation of genomic information with the medical record both for diagnostic purposes and for the data mining that will be available from the resulting combined databases, both clinical and genomic.
The present overly cautious approach to bio-molecular testing -- that is testing for suspected variants on the basis of clinical suspicion -- limits the science of data mining. With this limitation, any additional abnormalities are viewed as incidental to a skewed population of suspected cases.
Studies based on massive insurance data are flawed from the beginning. Insurance diagnosis is entered according to the studies the clinician wishes to do in order to meet the diagnostic requirements for the test -- representing tentative diagnosis, a speculative or at best a differential diagnosis. Often-times the insurance clerk picks a diagnosis that approximates the physician’s but which better meets the requirements for reimbursement.
The screening of all patients -- with their permission -- on the other hand, will build a correlation between molecular and clinical findings that is more valid for research and can more cleanly contribute to a differential diagnosis and differential risk factors.
We have only scratched the surface of the information that will be forthcoming from molecular biology. We need to open the flood-gate of genomic information for clinical correlation with statistical analysis on an ongoing real-time basis. The data from multiple clinics and private practices could be combined anonymously and indefinitely in a medical school or trusted institution; its value will grow with time. It’s diagnostic and correlative legs will reinforce one another with time.
When I sold my practice/clinic there were over fifty thousand charts, a gold mine for data mining. Most old medical records are ultimately neglected or lost as were mine. This is front line data, which differs from the national or regional data collected by institutions. It is local data and as such more personal and thus more explicit and if analyzed real-time, more critical.
This process of learning as we go is the scientific method and the basis of medical science. The learning process should extend to all physicians willing to include it in their practice. I feel passionate about working in a team setting, to bring genetic screening to the medical clinic, the private practice and the community.
The goal would be the free distribution of a standalone software application that would securely store the whole genome of patients, whilst correlating it with that patient’s real-time clinical data. More relevant, however, the continuing growth in data-mining in the two matched databases will lead to new insights and certainties not otherwise rapidly achievable.
As an aside one might observe that evidence based protocols place more emphasis on treatment than on diagnosis. Missed and wrong diagnoses remain a major problem, a fact partially borne out by the US’s poor ranking in the Global Burden of Disease (GBD) studies. A byproduct of this initiative will lead to enhanced diagnostic decision support and a quantum improvement in patient care and outcome.

Tuesday, November 11, 2014

Lung-Cancer Screening with Low-Dose CT


Share | Numerous journals weighed in on the pros and cons of CT screening for lung cancer in high risk patients. The idea of a series of three CT scans seems overly expensive and an over dose of radiation considering the number of false positives. A review in the Resident e-Bulletin of the NEJM the teaching topic reports the article by M.K. Gould in the November 6th issue: NEJM, 371, 1813-1820 outlines the following.
Lung cancer has an 18% 5 year survival rate and early detection would help. The National Lung Screening Trial (NLST) consisted of 50,000 patients from 33 centers. Low-dose CT was compared with chest X-ray reporting 20% fewer deaths in the CT group, 247 vs. 309 for 2 year follow up. If valid statistically, that would be 3 deaths per 1000 saved.
The screening produced 39% positive reports, 95% of which proved false, however. The author suggested that the resulting additional CT scans and invasive procedures produced few complications: 2% from needle biopsy, 4% by bronchoscopy, and 4% from surgery. 73% of the needle biopsies and bronchoscopies were negative and 24% of the surgeries were benign. Only 1% of these invasive procedures experienced complications, 20% of whom did not have lung cancer. (approximately 0.2% complication rate for non-cancer)
Twenty years ago, our diagnostic routine (Swedish Hospital IM Denver) for 30 and more pack a day smokers with a cough was bronchoscopy, bronchial brushings, culture and chest X-ray. With the low complication rate for bronchoscopy in non malignant patients, why would one choose the overly expensive three CT screening with it's radiation exposure and 95% false positive rate when a low risk and relatively inexpensive bronchoscopy yields more definitive results? It does become a question of sensitivity, however. One would have to re-examine the claimed 20% reduction in 2 year mortality and apply the same if not better trial for the sensitivity of fiber-optic bronchoscopy and bronchial brushings in the early detection of bronchial genic carcinoma. One would be dependent on conventional PA and Lateral for the detection of  non bronchial genic CA.

Saturday, November 8, 2014

Autopsy


Share | A sign hanging over the door to the pathology lab in Paris read, "Death comes to the aid of the living." Around 1793, Marie Francois Bichat (1771-1802) moved to Paris from Lyon bringing clinical pathology to the bedside. The formalization of autopsy as part of clinical medicine catapulted the art of diagnosis and pathophysiology into a science and onto a quantum advancement in the understanding of disease and the rationality of patient care.
Today that tool, autopsy, which made American medicine the science that it became is all but lost to physicians. We still claim to be the most advanced, but the numbers say otherwise. See global burden of disease (GBD).

Sunday, November 2, 2014

Genomic Medicine


Share | The Ambrose Monell Fellowship at Cleveland Clinic suggests, importantly, that the successful fellow will understand how to form hypotheses from clinical observations and to design experiments to effectively answer the hypothesis. In essence, this appears to be the near universal role of clinical genetics, working from the clinical to the laboratory.
Arguably, the opposite might yield greater diagnostic accuracy and efficacy by correlating the genomic profile back to the signs, symptoms and anthropomorphic data of the patient. Testing for only suspected abnormalities to confirm a diagnosis, removes the chance for finding a missed diagnosis, questioning a wrong diagnosis or establishing an individual patient database from which future information might be derived.
I'm looking for a place where I can link the patient history with an inexpensive, less than $100 US, genomic profile and apply statistical analysis to the associations -- translational medicine. The predictive accuracy will build over time, but it will not build until you have the data and the data must be personal or highly regional. The data must apply to the patient or patients in question. A national database fails in that regard. A clinician who keeps good records and for the life of his or her practice, will have a database of some fifty thousand records. These numbers may be adequate for good data mining, but your patient is a statistic of N=1.