The concept underlying outcomes measurement or 'predictive medicine' is a new, emerging discipline in the world of medicine. Trial and error treatment selection is wasteful. It costs patients financially and prolongs the hardship each patient and his or her family must endure. The success of a treatment consisting partly or wholly of prescribed drugs depends largely on individual patient physiology. Because drug treatments have different effects on different individuals, even the most informed and experienced doctors are often limited to their best guess. A software model such as Rx-D has high potential for success today due to the capacity of desktop computers to handle complex algorithms, and as a result of the practical success of the Internet, intranet, and extranet information systems.

A method that reliably selects a successful first treatment would greatly benefit patients, doctors, and HMOs alike.

TECHNOLOGY

Predictive Medicine's Rx-D product determines an expected outcome of each individual patient's response to a drug treatment by first compiling a patient profile. Next, the computer searches a specialized database for patients with similar profiles who previously received one of several treatments. Each treatment has a predicted outcome for a particular patient's profile. Results show that this virtual testing compared to present traditional methods significantly increases the likelihood of choosing an effective drug treatment on initial evaluation. (See Figure 1.)

Rx-D is based on work conducted by Dr. Joanne Luciano. Her model for final treatment response uses neural networks (or quadratic regression) that have been trained on clinical data.

The model for temporal response uses coefficients in a set of coupled differential equations that have been fitted to clinical data. The reason for combining these models with other types of models is to search for fundamental relationships among the pretreatment data that will extend what current models are able to learn from clinical data.