Flexion Is Flexible With The Data, Rigid With The Price

About: Flexion Therapeutics, Inc. (FLXN)
by: Mako Research

Flexion's only approved drug is clearly not superior to current treatments, in our assessment.

Flexion's only approved drug costs 57x the price of the brand-name immediate release formulation that has been in use for decades.

Recently, the SEC charged Flexion's Chairman, Patrick Mahaffy, with misleading investors about how well Clovis' flagship lung cancer drug worked compared to another drug resulting in a $20 million settlement.

Update 10/23/2018: I have edited one sentence in the the article to reflect that longer-term data is available in the supplemental materials. Note that the longer-term data show poor efficacy for Zilretta and the article has been updated to acknowledge the data exists.


Flexion (FLXN) has developed and won approval for ZILRETTA (Triamcinolone acetonide extended-release, TA-ER) for osteoarthritis-related knee pain; Zilretta is Flexion’s only approved product. Outside of Zilretta, the pipeline for Flexion is essentially barren; the company’s valuation relies on sales and commercialization of Zilretta. The company’s strategy to sell Zilretta is to incentivize physicians to push Zilretta to patients which would cause Medicare to overpay (by a magnitude of 57x) for a drug which I see as no more effective than current treatments and has a worse side effect profile. Flexion’s market capitalization is ~$700 million, with an enterprise value of ~$515 million which leaves little room for error especially when considering Flexion carries a substantial amount of debt.


Zilretta is not superior to current treatments for pain management, but is priced like it is. Zilretta is an extended release formulation of triamcinolone acetonide, a glucocorticoid used for intra-articular injections to control osteoarthritis-related pain. In the pivotal phase III trial studying osteoarthritis knee pain, Zilretta was superior to placebo in pain management, but was not better than the immediate release formulation that has been in use for decades (see chart below). Despite this non-superiority, Zilretta was launched at the princely sum of $570/dose, or about 57 times the price of the brand-name immediate release formulation. Flexion has decided to go the time-honored route of targeting Medicare and Medicaid (accounting for 60% of the injectable steroid market)1 by setting the price point outrageously high and employing a team of 100 “Musculoskeletal Business Managers” to educate physicians into paying 57 times what the competition charges, for no extra benefit.

Source: Company Presentation

Because of the lack of differentiation on efficacy, these Musculoskeletal Business Managers will have to find other ways to market Zilretta. One possible angle could be in superior glycemic control. It has previously been demonstrated that intra-articular injections of glucocorticoids can raise blood glucose in type 2 diabetic patients. One way Zilretta may be differentiated from the competition is by increasing blood glucose to a lesser extent than control steroids in diabetic patients. Recently, data were reported on this hypothesis 2: patients with long-standing T2DM and knee arthritis were allocated to a Zilretta or immediate-release intra-articular steroid injection. Blood glucose was continuously monitored for 7 days pre-treatment and two weeks posttreatment. As hypothesized, the extended-release formulation did not increase blood glucose to the same extent as the immediate-release formula in the first 3 days after the injection (see below). The average difference in continuously monitored blood glucose over these three days was reported as 19.2 mg/dL between groups, with Zilretta-treated patients having significantly lower blood glucose.

Source: Figure 2, Russell, Steven J., et al. "Triamcinolone acetonide extended-release in patients with osteoarthritis and type 2 diabetes: a randomized, phase 2 study." Rheumatology (2018).2

This difference in blood glucose is not meaningful. A difference of 19.2 mg/dL implies a total blood glucose change of 960 mg glucos3. This is approximately equal to 1.6 M&Ms (individual candies, not bags), ¼ of a Dum Dum sucker, or the amount of glucose burned during a (leisurely) one-minute stroll, distributed over the entire blood volume. In more scientific terms, Zilretta would result in an HbA1c that is approximately 0.03% lower than immediate-release;4 this is clinically meaningless in the management of type 2 diabetes, and is equivalent to taking off your socks and t-shirt before weighing yourself at the doctors’ office.5 In fact, a difference in HbA1c of 0.03% is less than the internal variation in HbA1c laboratory assays (0.2 – 0.3%).6

The difference in blood glucose is real...but does not favor Zilretta. Unfortunately for Zilretta, a deeper dive into the glucose monitoring data paints a disastrous picture. First, in our opinion the study was biased for success: the baseline blood glucose of the Zilretta-treated patients was meaningfully lower (p = 0.13, Table S2). Second, the company chose to cut-off the presentation of data 72 hours post-injection throughout the publication, while data was collected for two weeks following the injection. If we correct for both issues, the data show an absolute victor in the head-to-head comparison: immediate-release steroids.7

Source: Supplementary Table S2, adapted from Russell, Steven J., et al. "Triamcinolone acetonide extended-release in patients with osteoarthritis and type 2 diabetes: a randomized, phase 2 study." Rheumatology (2018).

Data adapted as follows: Baseline CGMG AUE values were divided by 7 to calculate average daily CGMG during the lead-in (the text incorrectly states that baseline values are from hours -72 to -1). “Day 1” was subtracted from “Days 1-2” to determine Day 2 CGMG AUE. “Day 1-2” was subtracted from “Days 1-3” to determine Day 3 CGMG AUE. “Day 1-3” was subtracted from “Days 1-7” to determine Day 4-7 CGMG AUE. “Day 1-7” was subtracted from “Days 1-15” to determine Day 8-15 CGMG AUE. These values were then divided by the baseline value to determine the change from baseline in CGMG, which is presented above.

Rather than reducing the glucose excursion, Zilretta extends and increases the glucose excursion for multiple weeks following the injection due to its extended release action. Flexion primarily discusses the first 3 days of data in the paper, which shows a very marginal benefit for Zilretta, though they do present longer-term data in the supplemental materials which shows that Zilretta had a worse glucose score than generic in the Days 1-7 and Days 1-15 periods. Although it’s impossible to statistically analyze this phenomenon without the raw data, it is clear that the large effects on blood glucose from Day 3 – Day 15 greatly outweigh the minor effects from Day 0 – Day 3; if any effect is noticed by patients or practitioners on HbA1c and daily glucose control, it will be that Zilretta is in fact worse than immediate release triamcinolone acetonide for weeks after the injection, not better. This appears to be in direct opposition to what the company reported in its scientific publication and press releases; this data will find its way into physician “education” materials that Flexion’s horde of sales representatives are plying physicians with to increase sales of their egregiously over-priced formulation at the cost of the American tax-payer and patients’ health.

Investors are likely completely unaware or overlooking recent developments that the Chairman of the Board at Flexion, Patrick Mahaffy, just settled a complaint with the SEC in his role as CEO of Clovis Oncology:

Source: SEC

Insider Sales

Since the public release of the Zilretta diabetes data on November 1, 2016, insiders have been sellers of stock. Former Chief Medical Officer (current CSO) Bodick Neil – has sold more than $2.4M in stock, starting almost immediately after the data was released and progressing at an unrelenting pace:

Source: SEC Filings

Former CFO Frederick Driscoll also sold stock as soon as his options vested, resulting in total sales of just over $736k since the diabetes data were released. This despite his already miniscule holdings in the company: Driscoll never allowed his equity stake in the company to rise above $200k, another signal that investors appear to be overlooking or ignoring.

As can be predicted from Zilretta’s high price, non-superiority to alternatives, and dubious side effect profile in diabetics, sales of Zilretta have been slow to take off: following the launch of Zilretta in October 2017, sales were $2.2M in Q1 2018 and $3.8M in Q2 2018, whereas analysts expect sales to reach an estimated $268M in FY 2020.8


Flexion’s valuation is wholly dependent on sales of Zilretta; the primary points of differentiation for Zilretta vs previous immediate release formulations are its 57x higher price and worse long-term effects on glucose in diabetics. Other than these, Zilretta is very similar to immediate release triamcinolone acetonide. If the sales proposition is one of value, Zilretta fails due to its extraordinarily high price. If the sales proposition is one of superior efficacy, Zilretta fails due to its non-superiority. If the sales proposition is one of improved glycemia, Zilretta fails due to its dubious benefit. We see no benefit of Zilretta over currently-offered alternatives, and the market capitalization of the company (~$700M) does not reflect this reality today.


1 Company presentation, March 6 2018.

2 Russell, Steven J., et al. "Triamcinolone acetonide extended-release in patients with osteoarthritis and type 2 diabetes: a randomized, phase 2 study." Rheumatology (2018).

3 19.2 mg/dL * 50 dL total blood volume = 960 mg glucose

4 If we consider a difference of 19.2 mg/dL in the first three days of treatment, and conservatively assume that HbA1c only reflects the last two month of blood glucose: (19.2 mg/dL * 3 days / 60 days) = 0.96 mg/dL increase in average monthly blood glucose; this is an HbA1c change of approximately 0.03%.

5 (Socks (100g) + t-shirt (150g)) / body weight (70 kg) = 0.36%; (HbA1c change of 0.033%) / (HbA1c of 8%) = 0.41%

6 Heinemann, Lutz, and Guido Freckmann. "Quality of HbA1c measurement in the practice: the German perspective." Journal of diabetes science and technology 9.3 (2015): 687-695.

7 Data from Table S2 were adapted to daily values by dividing by number of days in the time period of note resulting in the following table. Percent changes off baseline were then calculated for each group.

8 Composition of estimates from analysts at Wells Fargo, Northland Securities, Janney Montgomery Scott, Laidlaw & Company, BMO Capital Markets, and RBC Capital Markets

Disclosure: I am/we are short FLXN. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.