Trius Therapeutics (TSRX) is a small capitalization biotech company developing novel antibiotics. Its lead compound is tedizolid, which is a second generation oxazolidinone in Phase III trials for the treatment of acute bacterial skin and skin structure infections. The company already reported a successful Phase III trial in late 2011 and expects results from the second Phase III trial in early 2013. The success of the first Phase III trial takes a lot (but of course not all) of the clinical risk out of the upcoming Phase III results. Trius, however, still only trades at a market capitalization of around $190 million (with $70 million in cash). What then explains the relatively low market capitalization for a company with drug that has a good chance to be approved in 2014? Given the previously successful trial, it seems more likely that investors are more concerned about the potential commercial success of tedizolid than clinical trial success.
There are many ways to estimate the potential commercial path of tedizolid. One could make assumptions as to the target market size, share of that market it would capture, and pricing of the drug. Taking those assumptions out a number of years would allow the calculation of a net present value. While this is useful, it ignores what we already know about the market and is clearly only as good as the assumptions one makes. An alternative is to examine the performance of recently approved drugs to search for patterns in recent drug launches. If there is a large enough sample of launches, then one could base their estimates on a larger statistical analysis. Figure 1 provides the quarterly sales of the branded antibiotics that would most directly compete with tedizolid. These data were taken from the quarterly filing of the respective companies [Pfizer (NYSE:PFE) and Pharmacia for Zyvox, Viropharma (VPHM) for Vancocin, Cubist (NASDAQ:CBST) for daptomycin, Pfizer and Wyeth for Tygacil, and Forest Laboratories (NYSE:FRX) for Teflaro]. While there are certainly more drugs in this market place, these account for the branded drugs that have recently entered the market (Vancocin being a branded generic). There are a couple of interesting trends that seem to jump out. First, Zyvox is the clear winner of the group. Second, aside from Vancocin where additional generic competition entered in 2012, sales generally grow quite steadily across time. Third, each new successive drug launch (excluding the branded generic Vancocin) seems to have a lower trajectory than earlier drugs, although it is difficult to tell for sure because some drugs are still growing (daptomycin and Teflaro, for instance). So what does this mean for tedizolid?
While one could eyeball the chart and estimate potential tedizolid sales, there is enough data to be more systematic. In particular, I generated a statistical model of these sales figures and then used those results to create an estimate for tedizolid. The data set ended up having 143 observations and the results were fairly consistent across various models and estimation techniques. As one would have expected, the models showed that the most important correlate of current sales were the lagged quarterly sales (only the two previous quarters were statistically significant). To capture the clearly different trajectories of the drugs, I also include a series of dummy variables for the separate drugs (Zyvox was the excluded category for the statistical analysis). The results indicated that Vancocin, Tygacil, and Teflaro sales were lower (statistically significant) than Zyvox (daptomycin coefficient was negative but not statistically significant). The model also included a variable for the number of drugs on the market and this coefficient was positive but not statistically significant. This might seem odd given that figure 1 seems to show that newer drugs are doing worse, which implies that launching a drug into a more crowded market leads to lower sales. The statistical model did not confirm this expectation. The most likely explanation for the positive coefficient is the inclusion of the drug specific dummy variables (when they are excluded the market variable is negative but still not statistically significant), which is attributing some of the crowded market variance into the drug specific dummy variables. Finally, the model contains a variable that is the number of quarters a drug has been on the market in an attempt to capture a peaking process and this variable is negative but not statistically significant. In general, the model provided an excellent fit to the data with an r-squared of 0.9764, which roughly means the model is explaining over 97% of the variance in quarterly sales.
While the model does a good job at estimating sales, it is does not provide an estimate for the initial two quarters. Tygacil and Teflaro are the two most recent drug launches and likely offer the best estimates as to the first two quarters of tedizolid sales. In addition, they provide two different sales trajectories in that the models are much more bullish on Tygacil than Teflaro (perhaps because of the earlier nature of the Teflaro launch). Figure 2 provides two separate sales estimates for Tedizolid. The Teflaro assumption is that tedizolid will have a similar trajectory to Teflaro but its first two quarters of sales are 10% lower than those seen by Teflaro (lower to account for slower launches in a crowded market). The Tygacil assumption is the same in that tedizolid has a Tygacil trajectory and first two quarters of sales that are 10% lower than those seen by Tygacil. These assumptions might both end up being conservative but it is often useful to value a company based on low expectations and leave upside.
Figure 2 provides a number of interesting insights. First, even though the initial sales assumptions are pretty close (only a $6 million difference in the first two quarters combined), they produce two very different peak sales. Of course, this is most likely attributed to how negatively the model views Teflaro and its potential sales rather than the sales in the first two months. In fact, if you use the Teflaro beginning sales with the Tygacil growth profile, the peak sales end up being about $1 million less the Tygacil assumption in figure 2. Second, and in terms of trajectory, the time to peak sales is quite different between the models (19 months in Teflaro assumption versus 33 months in Tygacil assumption). Finally, the Teflaro model essentially expects tedizolid to have about 10 years on the market before having zero sales as opposed to the Tygacil model that shows sales only slightly below peak 10 years after launch. While the statistical model and sales estimates are interesting, what do they models tell us about the valuation of Trius?
In order to generate a net present value, I assume that the net profit margin is 25%, they launch in 6 quarters, and the discounting rate is 2.5% per quarter (roughly 10% a year). With those assumptions and the Teflaro sale trajectory, the Tedizolid NPV is about $77.15 million and with $71 million in cash it implies a $147 million valuation for Trius. The Tygacil assumption puts the Tedizolid NPV at $254.78 million and Trius at $324.78. Obviously this is a large difference, which is reflected in the very different sales estimates. At a current market capitalization of $191 million for Trius, it seems like the market is projecting a sales trajectory much closer to Teflaro than Tygacil. This seems excessively pessimistic in that outside of Vancocin with the generic competition, none of the previous drugs saw a dramatic sales decline. As such the shelf life of Tedizolid shown in the Teflaro model is likely close to a "most bearish" set of assumptions. Given that the current market capitalization of Trius is close to the "most bearish" set of assumption it leaves a lot more upside than downside, which would imply that the risk/reward profile at these price levels favors the bullish side.
Of course, any modeling exercise is a simplification, which is certainly the case with these estimates. I have not included potential partnerships, which would increase current cash but lower future profits that go to Trius. In addition, I have not explicitly modeled a changing dynamic, i.e. the statistical model assumes that the future correlates of sales will be the same as the past correlates of sales. Perhaps going forward the launch of a generic Zyvox or a greater attention to cost savings will dramatically alter the landscape. While that may occur, it is much more common that those expecting this time to be significantly different than the past are more wrong than right. In addition, the models might already have seen a changing dynamic with the repeatedly less successful launches. When all is said and done, the statistical modeling offers a unique snapshot of the past market and potential market for Tedizolid. The results imply that the market is pricing in close to a worst case commercial scenario, which leaves significant upside to current investors. The next set of steps would be to supplement this sort of statistical modeling with a deeper comparison of Tedizolid treatment profile with its future competition, which should allow us to better determine which path is most likely.