In the current issue of Value In Health, Weinstein and colleagues provide an excellent overview of mathematical models and their use in policy decisions [1]. Some relevant collateral examples are provided in environment protection policies, defense strategy, and health care. Key observations drawn from these examples are used to support an analytic framework for evaluating models and their use under Section 114 of the Food and Drug Administration Modernization Act.

The need for models to inform health-care decisions is compelling. Most interesting is the concept of “value of information.” It may be too costly to wait for years expecting long-term data to validate a model of economic value in health care. A key issue concerns the nature of the information that is needed to inform the decision, which is fundamentally related to its value.

These observations are especially relevant to emerging health-care technologies. A hot topic in some areas is the need for data from large, lengthy “real-world” studies before new products are adopted. The hope is that these ambitious studies could provide definitive evidence of economic value and bridge the gap between clinical trials and everyday care. But the interest in this work should be tempered by attention to the immediate needs of patients and the potential consequences of delaying decisions.

Another theme is the manner in which models are used and presented. Models are inherently limited in their ability to predict the future, but they can still be helpful. Certainly by organizing information, bringing to light key assumptions, and providing rigor to the review of a new technology, a model can be very useful. So perhaps some models could be presented to formulary committees as long as they are presented in a responsible fashion, with clearly stated assumptions, and cautious or conditional statements instead of inappropriate claims. In light of these observations, is Section 114 of the Food and Drug Administration too restrictive? It is not entirely clear. We still need more experience in seeing how its application evolves to understand its full implication.

A key concern about Section 114 is its requirement that economic information relate directly to approved indications. In the case of a medication indicated for the prevention of myocardial infarction and stroke, will a model that projects life years saved based on the survival impact of these clinical events be interpreted as relating “directly” to an approved indication? Actions by the FDA to date suggest that projection of life years gained from event-level evidence is acceptable.

Under Section 114, a projection of life years saved for a product indicated only to lower cholesterol may not be appropriate. However, the regulation might already allow for presentation of a cholesterol-to-life expectancy model, provided that it does not incorporate the product being reviewed by the formulary committee. That is certainly one way to envision a nonpredictive use: the model is presented to understand the overall costs and consequences of hypercholesterolemia, without new products plugged in.

Finally, the role of Section 114 should be considered in terms of formulary committee demands. Right now, some of these committees seem more interested in simple and straightforward information, often well within indicated uses. They may approach complex models and end points like quality-adjusted life years with skepticism. Their demand for information will change as they gain more experience.

The proposals raised by Weinstein and colleagues have a lot of merit. The “value of information” needs to be considered when we interpret the role of mathematical models. The examples from other areas, such as environmental policy, are extremely informative. With respect to Section 114, we need to see how its implementation evolves over time.

References
1 Weinstein MC, Toy EL, Sandberg EA, et al. Modeling for health care and other policy decisions: uses, roles, and validity. Value Health 2001;4:(5).