Feedback in Epidemiology for Market Sizing

In early 1995 data showed that the HIV/AIDS population was growing at a terrifying rate in the United States. Moreover, although the so-called AIDS cocktail was in clinical trials, very few people were aware of what its performance would turn out to be. Ventana was engaged by a pharmaceutical company to develop a decision-support tool for managing the company’s launch of a new AIDS drug. This tool used a dynamic simulation model to predict the number of HIV and AIDS cases over time.

The predictions did not square well with the company’s internal assessments. Indeed, the model generated future sales that were an order of magnitude lower than the most likely sales forecast used by the business unit that was developing the drug. Initially, this did not create much of a stir. The idea was to get a model ‘structure’ together quickly (within a few weeks) so that the business unit itself could select appropriate inputs. However, the model did incorporate a simple epidemiological model that was calibrated using the best data available from the CDC at the time. Moreover, in processing the data, serious attention was devoted to addressing systemic data biases introduced by the data reporting methods. In the context of the model, the CDC data clearly suggested that the business unit’s future percentage growth rate for the infected population was far too high. While the positive feedback inherent to an infectious disease was accelerating the rate of increase in AIDS cases, an important negative feedback mechanism was emerging in the most important high risk sub-populations. The model predicted this negative feedback would strongly suppress future growth and the rise in AIDS cases would soon level off. The specific causal driver of the negative feedback was inferred to be a logical response of well-educated subpopulations at risk, i.e. the use of safer sexual practices.
Ventana’s finding about the epidemiology led to formal peer review involving company scientists, the marketing staff and Ventana. The review was at times very contentious, but after a few months and more detailed modeling, the Ventana findings were confirmed. In addition, the peer review process added a great deal of technical richness and understanding to the epidemiological issues, including the realization that the requirements of public policy can sometimes make official statements about epidemiological data difficult to interpret from a scientific standpoint.

Many people were caught by surprise by the outcome of the peer review. The quick, simple dynamic model of the disease proved more robust than people expected. Moreover, the coherence of the epidemiological story was unnerving from a business standpoint. Neither company scientists nor marketers now supported the original market size estimate, which served as a key basis for the new drug development program.

The Ventana epidemiological model proved accurate, including the near-term flattening in the number of AIDS cases. Historical growth had been exponential. Because of the sunk cost, the new AIDS drug was launched. It became a useful niche product, but it was not the blockbuster on which the company was counting. The senior management team of the weakened company was replaced, and the company was soon acquired by a competitor.