Fixed dose combinations forecasting
Gary Johnson, Managing Director, Inpharmation
This is the last talk of the day, and Gary doesn’t have much time. He insists he will speed through his talk, getting straight to the meat. His main messages are:
- Keep things simple.
- Simple models produce better results than judgement alone. We should look for models and simple patterns to help us generate accurate results.
We in the pharma industry can spend hours, days, weeks, even months coming up with highly complex formulae; we generate graphs, pie charts and vend diagrams. But sometimes, we can be too clever.
Gary says we’re in danger of making our analysis of the market more complex than it needs to be the doctors do. Doctors don’t look at pie charts before they prescribe. They examine the individual patients and make a split-second decision. They care about control the disease/ symptoms as quickly as possible.
He gives the example of Rube Goldberg’s contractions where
this over complicating processes is taken to a comic extreme.
Search for simple models in market data; these tend to produce better results than judgement alone. But be careful not to make analysis too complex.
Logic trap
Analogy: Gary says that when his wife goes away, he starts reverting to student behaviour, eating baked beans and sausages. He goes to buy these at the supermarket and sees a can of baked beans with sausages. He’s tempted. There are two different brands on offer: Heinz and Brandson. Since he usually buys the Heinz baked beans, he naturally chooses that brand for his bean and sausage combination.
Test a hypothesis: examine the data
Are doctors persuaded by brand when it comes to FDCs? It’s a very simple hypothesis that can be easily tested just by examining the data.
Gary brings up some graphs; we can see that the FDC behaves like a brand rather than like a new product. The relative market shares of the parent brand and the combination product are similar.
Once you have a hypothesis, produce a model.
He shows a graph showing 3 brands with FDCs. In all 3 cases, the proportion of sales of the parent and combination brands is the same. He examines how the sales of the new combination drug impacts that of the parent drug. In the US, FDCs do eat into the sales of the existing product (and they do so proportionally). This is known as the ‘law of fair shares’.
But this isn’t universal; in Germany, FDCs are very popular and the market expands when they’re introduced (i.e. sales of the product drug aren’t so negatively impacted). We should look out for these variations.
Final remarks:
Coming up with highly complicated, sophisticated models is exhausting. And tired forecasters are far more likely to make mistakes.
Look for simple patterns in data and you can come up with effective models.
