Bayesian Statistics are in many cases perfectly suited for medical devices as opposed to clinical trials using Gaussian distribution.


We use a single example to explain (1), the Likelihood Principle, (2) Bayesian statistics, and (3) why classical statistics cannot be used to compare hypotheses.


A Bayesian analysis uses the posterior distribution to summarize the state of our knowledge. The posterior distribution combines information from the data at hand expressed through the likelihood function, with other information expressed through the prior distribution.

Refer to What is Bayesian statistics and
why everythine else is wrong.
Michael Levine

Bayesian statistics not Gaussian distribution science

Back in 1997 as part of an FDA mandate from Congress - the use of Bayesian statistics was both endorsed and promoted by the FDA. Bayesian statistics is better suited to this medical device - than the overly encumbered pharmaceutical model.


Proper science requires that the investigation be based on an applicable risk model. There are:

• No issues of biological diversity were associated with his treatment
• No issues surrounding a new chemical metabolic functions

Therefore, the traditional prelude of animal studies and the Gaussian distribution curves having expanded populations is an inappropriate model. There are no unanticipated risks where large population groups were required to minimize unintended results.

Protocol Enrollment

What we propose is to create an enrollment strategy for simple clinical demonstration. This demonstration would enroll 13 patients for treatment. The treatment regiment would began by taking likely blood parameters be a treatment every 12 hours apart for four or five days. Data of blood chemistry parameters can be analyzed before and after the treatments will demonstrate efficacy. After five days, the data would demonstrate:

• No harm had come to the patient
• No interference with their existing treatments have occurred
• No observable progression of their disease has occurred (other than a projected)

Some of those 13 patients would show improvements.

Typical Protocol

If 13 patients are selected to a protocol and treated with consistent results - then these results can be projected into larger groups.


We must evaluate what patient parameters are most compelling. The treatment regiment is simply - routine treatments every period for a specified routine. After a completed period results can be measured. Not every patient responds. The other 9 patients in the group may not respond - but those who do not respond - are not harmed by our treatment. The 4 patients who do respond will swear it is a miracle. Some follow up treatment sessions are required to achieve complete reversal and full recovery.

Data Correlation

With established facts, using Bayesian statistics, predicted out comes can be predicted into larger groups with statistical reliability. If any unintended consequences occur. it is likely that the consequences resulted from a unanticipated definition in the enrollment definition. This is the proper science that should be applied to the clinical demonstration for medical devices.


Compliance Consultants has been consulted on clinical trials for many Class II, non-significant risk devices.