Kymera Therapeutics (KYMR) spent its first five years as a public company as a platform story: a protein-degrader engine, a Sanofi partnership, and a market that admired the science without paying for it. It priced its IPO at $20 in August 2020 and was still trading near $21 two years later. At our $112.50 reference price it carries a $12.15B fully diluted equity value on a single Ph2b readout. BROADEN2, due by year-end, tests KT-621 in atopic dermatitis (AD), and the question is how much efficacy a pill can give up against injectable IL-4/IL-13 blockade and still win, and how much of a good answer the price already holds.
The re-rating came in steps. Over 2024 and 2025 management narrowed the company around KT-621, a once-daily oral degrader of STAT6, the transcription factor downstream of IL-4 and IL-13 that Dupixent (dupilumab) turned into a franchise. Healthy-volunteer data in Jun 2025 beat the company’s target profile. In December 2025 the open-label BroADen Ph1b showed a 63% mean EASI reduction at Day 29 in 22 patients, and the shares rose 42% on the day. Partners then paid for the platform: Gilead exercised its CDK2 option for $45M in Apr 2026, after a $40M upfront in Jun 2025, and Sanofi earned a $20M milestone on its IRAK4 degrader KT-485 in Jun. BROADEN2 itself is a randomized, double-blind Ph2b of ~200 adolescents and adults with moderate-to-severe AD, testing three doses against placebo over 16 weeks. Enrollment closed ~6 months early and the last Wk16 visit was completed in Oct.
We are constructive on KT-621 and cautious on the stock. On our central bridge from BroADen, the true placebo-adjusted Wk16 EASI effect is ~33pp, a touch below dupilumab’s 34pp to 36pp in Ph3. A ~50-patient arm will print within ~15pp either side of the truth, and readers will use the better of two doses. At $112.50, our reference price (a judgment based on the past month of trading), with $1,645.9M of net cash at 30 Sep (3Q 10-Q) and an EV of ~$10.50B, a dupilumab-like result (line A) is worth ~$102 a share on our inputs. It reaches $127 to $133 only if the AD franchise follows our extension of consensus revenue to 2038.
The harder question is the price. A dupilumab-like result explains $102 of the $112.50 reference price, and the last ~$10 has several explanations. It could be a larger AD franchise (~$6.9B of peak sales rather than our $6.1B, against Dupixent’s ~$8B to $10B of AD sales in 2025 on our estimate), surer approval, more value in asthma and other indications than we assign, or a weaker print paired with a large franchise. A takeout premium is one more we cannot price. BROADEN2 speaks most directly to approval odds and the size of the effect, and much less to share or to asthma. A printed 40pp or more is valued above the price, but only if both doses confirm it, and the Low scenario is worth $69.5. This note rebuilds the bridge from BroADen to Wk16, separates trial noise from true effect, tests how much a pill can concede, and back-solves the franchise the price requires. The closing box covers positioning.
Exhibit 1
A dupilumab-like result explains $102 of the $112.50 reference price. The last $10 has four different explanations.
Value per fully diluted share on Clinaptis line A (a BROADEN2 delta of 30 to under 40pp), built block by block, then zoomed in on the gap to the share price.
What a dupilumab-like readout is worth
Zoomed in on $80 to $120: four ways to reach $112.50, and one way it looks too high
Each row in the lower panel changes one thing and holds the rest at line A, so they are alternatives, not a sum. A price near $112.50 does not say which one the market is using. Bigger share or surer approval means paying for a larger AD franchise. The asthma route means paying for indications we value at about $8 a share. The line B route means a good-not-great print could still support the price if the franchise is large. BROADEN2 speaks most directly to approval odds and the size of the effect, and much less to share or to the value of asthma (data late 2027).
View data
| Block, line A | $M | $ / share |
|---|---|---|
| Atopic dermatitis (rNPV including terminal value) | 7,180 | 66.50 |
| Asthma | 387 | 3.58 |
| Other type 2 indications | 492 | 4.55 |
| IRF5 degrader KT-579 | 394 | 3.65 |
| Platform and partnered programs | 1,500 | 13.89 |
| Net cash, 30 Sep 2026 | 1,646 | 15.24 |
| Corporate cost, present value | −554 | −5.13 |
| Equity value, line A | 11,044 | 102.28 |
| Route to the price | Value tested, $ / share | Source |
|---|---|---|
| Peak share 23% (line A base 20%) | ~112.50 | A2b: line A needs ~23% share |
| Approval from Ph3 start 80% (base 70%) | 112.65 | A2b |
| Asthma and other type 2 at ~2.2x our value | ~112.50 | Approximation: scales the A+ read-across sensitivity (+$1.31 for +0.15 multiplier) |
| Line B, consensus-sized franchise | 112.25 to 117.65 | A4 (line B, shapes 1 and 2) |
| Line A, conservative franchise shape | 84.69 to 88.13 | A4 (shape 3) |
1. BroADen’s 63% mean is strong, but response depth and the blinded Wk16 effect are not established
BroADen cut mean EASI by 63% at Day 29, yet only 29% of patients reached EASI-75 and there was no placebo arm. Getting from that open-label number to a blinded Wk16 effect takes assumptions BroADen cannot identify. This section builds the bridge, and it lands at ~33pp within a range of 23.5pp to 42.3pp.
A strong mean, with no evidence yet of deeper responses
BroADen was an open-label Ph1b of KT-621 in 22 patients (10 on 100 mg, 12 on 200 mg, once daily for 28 days). At Day 29 the company reported a mean EASI reduction of 63%, with EASI-50 in 76%, EASI-75 in 29% and vIGA 0/1 in 19%. Median STAT6 degradation was 98% in blood and 94% in skin, with no serious adverse events and no conjunctivitis. The two doses were indistinguishable on efficacy (mean 62% and 63%), consistent with a saturated target.
Against dupilumab at the same time point, the mean looks better and the deep responses do not. In the pooled SOLO trials (n=457), Day 28 mean EASI improvement was 52%, with EASI-50 in 57% and EASI-75 in 28%. PROLEAD, an open-label real-world dupilumab cohort with a baseline EASI near BroADen’s (n=585), shows 52.5%, 62% and 34%. KT-621 is 10pp to 11pp ahead on the mean and 14pp to 19pp ahead on EASI-50, but level with SOLO on EASI-75 (29% against 28%) and 5pp behind PROLEAD.
That pattern fits a pill that moves more patients out of the bottom of the distribution (24% of KT-621 patients below EASI-50 against 43% on dupilumab in SOLO) without adding deep responders. We hold it as a hypothesis, not a finding. At 21 evaluable patients one response flip is 4.8pp, and open-label scoring, a lower baseline (EASI ~25 against ~33 in SOLO) and noise could explain part of the gap. In BROADEN2, EASI-75 and vIGA will show whether a favorable mean extends to deeper responses.
The Wk16 delta depends on the placebo arm as much as on the drug
Dupilumab sets the reference. Its Ph3 monotherapy trials (SOLO-1 and SOLO-2) showed placebo-adjusted Wk16 EASI improvements of 34pp and 36pp, and management expects BROADEN2 to land “in that ballpark of what dupilumab has shown at week 16” (Q2 2026 call, 5 Aug) without giving a numeric target. Management also argues that a safe, effective oral need not match it commercially. We take ~30pp as line A, the lower edge of a dupilumab-like result.
The risk is the placebo arm: a strong drug arm can still produce a disappointing delta if placebo does unusually well. Across 27 monotherapy trials with a 16-week primary endpoint and at least 20 placebo patients, median placebo improvement in EASI is 39% (interquartile range 26% to 45%; Exhibit 2), 42% in the 11 trials started since 2019 and 34% in the six large Ph3s. These are descriptive, not a predictive interval: the registry mixes analysis sets and rescue rules, and the SOLO available-data means can raise apparent placebo improvement. Dupilumab itself shows the risk: its Ph2b placebo arm improved only ~20%, so its Ph2b delta (~47pp to 55pp) overstated its Ph3 delta (~35pp).
Management says tighter trial controls have “accounted for” higher placebo rates (Stifel forum, 24 Sep), which we cannot verify, and none of our 27 trials exceeds 63%. We carry a working placebo range of −30% to −40%, centered on −35%, with a tail to −45%.
At a placebo response of −40%, clearing 30pp needs a raw drug arm of ~−70%, close to the ~−74% best IL-4/IL-13 monotherapy arm in our dataset, if the drug arm does not rise with placebo. Whether placebo and drug responses rise together is the open question, and we label that link κ. If a higher placebo response lifts the drug arm one-for-one (κ=1), the delta is unchanged. If the drug arm does not move (κ=0), the delta falls one-for-one, so a 5pp swing in placebo response costs 5pp, as much as the width of line B. Dupilumab’s move from Ph2b to Ph3 looks like κ near 0 and upadacitinib’s near 1, and two programs cannot support a fitted value, so we run κ at 0, 0.5 and 1 and weight none.
Exhibit 2
Placebo arms run from 6% to 63%; a 30pp delta needs a drug arm of 60% to 75%
Placebo-arm EASI improvement in 27 atopic dermatitis trials (median 39.4%), and the drug-arm improvement needed for a 30pp or 40pp placebo-adjusted delta at placebo responses of 30% to 45%.
A 30pp delta needs a drug arm of 75% at a placebo response of −45%, the top of the observed range, and a 40pp delta needs 75% at −35%. The 27 values describe where past placebo arms landed across eras, populations and estimands, not a predictive interval, because a ~50-patient BROADEN2 placebo arm adds its own sampling error, and if the drug arm does not rise with placebo (κ=0) a placebo draw in the tail moves the delta one-for-one.
View data
| Placebo response | Drug arm for 30pp | Drug arm for 40pp |
|---|---|---|
| −30% | 60% | 70% |
| −35% | 65% | 75% |
| −40% | 70% | 80% |
| −45% | 75% | 85% |
27-trial registry
| Trial (NCT) | Name | Start | Placebo n | Placebo improvement, % |
|---|---|---|---|---|
| NCT01979016 | 2013 | 27 | 5.8 | |
| NCT03703102 | 2018 | 57 | 15.0 | |
| NCT01859988 | 2013 | 61 | 20.2 | |
| NCT02925117 | 2016 | 39 | 23.0 | |
| NCT03054428 | 2017 | 85 | 23.6 | |
| NCT05732454 | 2023 | 28 | 24.6 | |
| NCT05911841 | 2023 | 35 | 25.2 | |
| NCT04146363 | 2019 | 141 | 26.0 | |
| NCT04178967 | ADvocate2 | 2019 | 146 | 28.0 |
| NCT05131477 | STREAM-AD | 2021 | 79 | 29.4 |
| NCT02277769 | SOLO 2 | 2014 | 105 | 33.7 |
| NCT03435081 | BREEZE-AD5 | 2018 | 28 | 34.1 |
| NCT03607422 | 2018 | 142 | 34.5 | |
| NCT03912259 | 2018 | 83 | 39.4 | |
| NCT02277743 | SOLO 1 | 2014 | 97 | 39.5 |
| NCT03569293 | Measure Up 1 | 2018 | 128 | 40.7 |
| NCT03443024 | 2018 | 52 | 41.1 | |
| NCT05923099 | 2023 | 52 | 41.7 | |
| NCT03568162 | 2018 | 42 | 42.8 | |
| NCT05999799 | AtDventure | 2023 | 31 | 44.5 |
| NCT02576938 | 2016 | 49 | 45.9 | |
| NCT03533751 | ATLAS | 2018 | 60 | 49.4 |
| NCT05469464 | 2022 | 55 | 50.4 | |
| NCT06101823 | 2023 | 23 | 51.0 | |
| NCT03517566 | 2018 | 72 | 55.0 | |
| NCT03747575 | 2019 | 32 | 58.2 | |
| NCT04800315 | 2021 | 56 | 62.6 |
Two population differences could complicate the comparison
BROADEN2 may not resemble the trials that anchor line A. SOLO enrolled patients with a mean baseline EASI of ~33 against ~25 in BroADen, and management says recent AD trials enroll in the mid-20s but will not say where BROADEN2 landed (Q2 2026 call, 5 Aug; Citi Back to School Summit, 10 Sep). Prior treatment is the second unknown: BROADEN2 has “no cap on prior biologics” for patients who stopped for reasons other than non-response, and most patients come from outside the US. We make no adjustment for either because the net direction is unclear to us, so baseline EASI and the prior-biologic share are the first two checks on release day.
The Central bridge reaches ~33pp, but the open-label haircut drives the answer
Our Central estimate for KT-621 is a 32.9pp placebo-adjusted Wk16 EASI improvement, a touch below dupilumab’s 34pp to 36pp in SOLO and inside line A. Getting there from BroADen’s 63% Day 29 mean takes an adjustment the Ph1b data cannot establish.
The bridge makes three assumptions for the middle dose M. We subtract a 15pp haircut for open-label scoring, patient selection and regression to the mean (range 0 to 25pp), subtract a 20% Wk4 placebo response, and divide by f=0.85, the share of the Wk16 effect already visible at Wk4 (comparator ranges in Appendix A1). That gives (63 − 15 − 20) ÷ 0.85 = 32.9pp.
The arithmetic is simple and the assumptions are not. BroADen identifies neither the haircut nor f, and the haircut carries most of the uncertainty: at f=0.85 each 5pp of haircut moves the answer by ~6pp, from ~51pp with no haircut to ~21pp at 25pp. This one assumption takes the same Ph1b result from line C to above A+ (heatmap, Appendix A1).
Our Low (23.5pp), Central (32.9pp) and High (42.3pp) scenarios illustrate that range; they are unweighted and not bounds. The High case needs a raw KT-621 arm improvement of 76% to 79% at a placebo response of 37%, above the ~75% best IL-4/IL-13 monotherapy arm, and we keep it as an upside sensitivity rather than exclude it on precedent. Management said it would not be surprised if the top two doses were “much closer to each other than the lower dose” (Stifel forum, 24 Sep), so the Central case puts M and H at ~33pp and the low dose L at ~8pp, 25% of the high-dose effect. The disclosed doses are not known.
Bottom lineBroADen supports a plausible dupilumab-like outcome but does not establish one. The ~33pp Central estimate is a reference point that depends on an unobservable haircut, not a calibrated forecast. The larger question is how reliably a ~50-patient arm will reveal the true effect, which is the subject of Section 2.
2. A 50-patient arm prints within ±15pp of its true effect, and the better of two doses flatters the result
Even if KT-621’s true Wk16 effect is ~33pp, BROADEN2 will not report exactly that. With ~50 patients per arm, sampling noise alone moves a drug-minus-placebo estimate by about ±15pp, and investors will trade the printed number. Section 1 asked what the effect is; this section asks what a trial this size would report.
Small arms give a wide range of prints
Patient-level dispersion of the % EASI change at Wk16 is rarely reported, so we proxy it: 38pp for the drug arms, from the lebrikizumab Ph2b least-squares mean SDs (37pp to 38pp in the drug arms, 57pp in placebo), a placebo SD of 1.0 and 1.5 times that, and a SOLO-like lower case (28pp drug, 34pp placebo). The SOLO registry values use available data and exclude rescued and missing patients, so they likely understate what a BROADEN2 estimand will carry (Appendix A1).
At 50 patients per arm the standard error of a drug-minus-placebo contrast is ~7.6pp, a 95% sampling range of ~±15pp (~±19pp at 1.5 times placebo dispersion, ~±12pp SOLO-like). Management describes the study as ~200 patients and says high-enrolling trials usually enroll slightly more than planned (Q2 2026 call, 5 Aug), so 50 is a floor. A drug with a true 33pp effect can therefore print in the 20s or the 40s without any change in efficacy; the range describes one estimate given an assumed true effect, not the drug’s efficacy.
The better of two doses adds an upward bias
Investors will anchor on the better of the two top doses. With one shared placebo the estimates are correlated, but picking the higher still puts the expected maximum of two equal-effect arms ~3pp above the truth. In the Central scenario either top dose prints 30pp or more 80% of the time against 65% for one dose (Exhibit 4 data table); under Low the better dose lands across the four lines almost at random (A or better 31%, B 28%, C 24%, D 18%). Sampling noise alone can therefore put the print in a different band from the true effect. Exhibit 3 shows how far. The better dose prints between 15pp and 38pp in 90% of Low readouts, 25pp to 48pp in Central and 34pp to 57pp in High, so a dupilumab-like 35pp print sits inside all three ranges and cannot, on its own, tell them apart.
Exhibit 3
A dupilumab-like 35pp print sits inside the 90% range of the Low, Central and High cases alike
Where the better of two doses is likely to print, by bridge case: placebo-adjusted Wk16 EASI improvement, percentage points. The three cases are unweighted, and prints run about 3pp above each true effect because the headline picks the better dose.
Drag across the chart to test a threshold from 20pp, the D/C line, upward.
Right-hand column: share of readouts that print 35pp or more.
At 35pp, reaching it is plausible in every case (11% to 93% of readouts), so it cannot tell the cases apart.
Vertical lines mark the Clinaptis readout lines (20, 25, 30 and 40pp; letters D to A+). Triangles mark each case’s true effect (Appendix A1 bridge: Low 23.5, Central 32.9, High 42.3pp). The cases rest on a haircut and a Week 4 share that no data identify, so they are scenarios, not forecasts. Dupilumab is a cross-trial benchmark from different populations.
View data
| Case | True effect, pp | 5th pct | 25th pct | Median print | 75th pct | 95th pct | Prints 30pp or more | Prints 35pp or more | Prints 40pp or more |
|---|---|---|---|---|---|---|---|---|---|
| Low | 23.5 | 15.1 | 21.9 | 26.5 | 31.3 | 38.1 | 30.8% | 11.3% | 2.8% |
| Central | 32.9 | 24.6 | 31.3 | 35.9 | 40.7 | 47.5 | 80.2% | 55.1% | 27.8% |
| High | 42.3 | 34.0 | 40.7 | 45.3 | 50.0 | 56.9 | 98.7% | 93.2% | 77.7% |
A 40pp print is not a 40pp drug
The gap widens above line A. At the Central true effect the better of two doses prints 40pp or more 28% of the time, but both doses do so only 7%, so about three in four A+ prints are one dose far above the other. One impressive estimate can move the stock, but it is weak evidence of a superior efficacy profile.
Pooling two equal-effect doses gives a standard error of ~6.6pp, so two observed 40pp doses would update to ~36pp to ~38pp under an illustrative prior around 33pp, and a 40pp and 30pp pair to ~34pp (Appendix A1; we do not claim the priors). A single 40pp estimate is insufficient to assign A+ economics, which is why Section 4 treats A+ as a reaction scenario and credits it only on confirmation.
Exhibit 4
In 100 readouts of a ~33pp drug, 28 show a dose at 40pp or more. Only 7 show two.
Simulated BROADEN2 readouts if the true placebo-adjusted EASI effect is 32.9pp (Central): two equal-effect top doses, ~50 patients per arm, one shared placebo. Each dot is one readout.
Three in four 40pp prints rest on a single dose. We credit A+ economics only when both doses confirm.
The same test at a lower or higher true effect
Share of readouts in which a dose prints 40pp or more. Low and High are contrast cases, not weighted.
Sampling probabilities given each true effect, not likelihoods of the scenarios. The true effect rests on a haircut and a Week 4 share that no data identify (Appendix A1). A single, pre-chosen dose prints 40pp or more 18% of the time at Central.
View data
| Scenario | True effect, pp | A+ (40+) | A (30 to <40) | 30pp or more | B | C | D | Clears first Holm step | 1.5x placebo SD: 30pp+ / A+ / Holm | SOLO-like SD (28/34pp): 30pp+ / A+ |
|---|---|---|---|---|---|---|---|---|---|---|
| Low | 23.5 | 2.7% | 27.9% | 30.6% | 27.7% | 24.1% | 17.5% | 88.4% | 35.1% / 7.2% / 64.0% | 22.9% / 0.7% |
| Central | 32.9 | 28.0% | 52.4% | 80.4% | 13.9% | 4.6% | 1.0% | 99.5% | 74.2% / 33.0% / 91.8% | 81.3% / 20.3% |
| High | 42.3 | 77.8% | 20.9% | 98.7% | 1.2% | 0.1% | 0.0% | 100.0% | 95.3% / 71.9% / 99.2% | 99.4% / 78.2% |
| Scenario | Single dose 30pp+ | Better of two 30pp+ | Both doses 30pp+ | Single dose 40pp+ | Better of two 40pp+ | Both doses 40pp+ |
|---|---|---|---|---|---|---|
| Low | 19.5% | 30.6% | 8.4% | 1.5% | 2.7% | 0.2% |
| Central | 65.1% | 80.4% | 49.7% | 17.6% | 28.0% | 7.3% |
| High | 94.7% | 98.7% | 90.8% | 61.9% | 77.8% | 46.1% |
Statistical significance is unlikely to be the swing factor
Against an illustrative first Holm step (z=2.394, standing in for the undisclosed multiplicity plan and stricter than nominal p<0.05), the best dose clears the threshold 99.5% of the time in Central, 88% in Low and ~100% in High. These are not forecasts of formal trial success, but they suggest the size and consistency of the point estimates, not the p-value, will drive the stock’s reaction.
All probabilities are conditional on the Section 1 true effect, under a normal approximation with shared placebo, 50 patients per arm and no dropout, rescue or estimand differences. They show how a true effect could print, not how likely it is, and we do not blend the scenarios. They also move with placebo: at κ=0, a placebo response of −45% rather than the −34% reference takes the Central 32.9pp to ~22pp, line C.
Our viewWe weight agreement between the two top doses over the best single estimate. A 40pp print from one dose and 30pp from the other updates only to ~34pp under our illustrative prior, so we credit A+ economics only when both doses confirm.
3. Patients will trade ~17pp of clear-skin response for an oral pill, and value itch relief even more
The best quantitative handle on how much efficacy KT-621 can concede to dupilumab is a patient-preference study: respondents accepted ~17pp less clear-skin response for an oral pill, but valued meaningful itch relief at about three times that.
In a 404-patient discrete-choice survey of adults in the UK, France and Spain, 68% with no self-injectable experience (Thomas et al., BMJ Open 2022), respondents accepted up to 16.6pp lower probability of clear or almost-clear skin (95% CI 9.2pp to 24.0pp) for a daily oral over a fortnightly injection, other attributes equal.
That is stated preference, not revealed uptake and not an efficacy floor. Mapping it onto dupilumab’s 27.6pp IGA 0/1 delta would require shared placebo response and endpoint definitions that cross-trial comparisons lack, so we do not convert it into a BROADEN2 acceptance line.
Exhibit 5
Patients in one stated-preference study traded ~17pp of IGA response for an oral; the comparator landscape spans 10pp to 51pp
The two panels are on different measures: the IGA 0/1 deltas the trials delivered, and the 16.6pp (95% CI 9.2pp to 24.0pp) that a 404-patient survey traded for an oral pill over an SC injection every 2 weeks, other attributes equal. Together they are not an acceptance line for KT-621.
The survey trade-off is not a threshold for approval or uptake. The same study values a 20% to 50% rise in itch response at 50.1pp and avoiding frequent monitoring at ~21pp; these are different attribute changes, not interchangeable effects. It prices no boxed-warning penalty beyond check-ups.
View data
| Agent (source trials) | IGA 0/1 Δ, pp | Dosing and safety features (descriptive) |
|---|---|---|
| Dupilumab 300 mg Q2W (SOLO-1/2 mono) | 27.6 | SC Q2W; no boxed warning; ocular warning; no lab monitoring |
| Lebrikizumab 250 mg Q2W (ADvocate-1/2 mono) | 26.4 | SC Q2W; no boxed warning; ocular warning; no lab monitoring |
| Tralokinumab 300 mg Q2W (ECZTRA-1/2 mono) | 10.0 | SC Q2W; no boxed warning; ocular warning; no lab monitoring |
| Nemolizumab 30 mg Q4W (ARCADIA-1/2, +TCS) | 11.4 | SC Q4W; no boxed warning; no lab monitoring |
| Upadacitinib 15 / 30 mg (Measure Up 1/2 mono) | 36.9 / 50.5 | Oral QD; boxed warning; lab monitoring |
| Abrocitinib 100 / 200 mg (JADE COMPARE, +TCS) | 21.9 / 34.6 | Oral QD; boxed warning; lab monitoring |
| APG777 720 mg (APEX Part A / Part B) | 17.6 / 35.1 | SC ~Q8-12W; conjunctivitis 10.6% (all terms, Part B) |
| Amlitelimab 250 mg Q4W (Ph2b) | 17.0 | SC Q4W; no boxed warning; no lab monitoring |
| Rocatinlimab 150 mg Q4W (Ph2b) | 17.4 | SC Q4W; no boxed warning; no lab monitoring |
| KT-621 (oral QD) | not yet reported | Oral QD; 0 conjunctivitis AEs in 22 patients over 28 days; no boxed warning or lab monitoring reported; safety not established |
Itch versus route. Itch outweighs route. Raising the probability of meaningful itch reduction at Wk16 from 20% to 50% was worth a 50.1pp loss in clear-skin probability (95% CI 38.5pp to 61.8pp), about three times the route trade-off. The attribute changes differ in size, so the ratio is directional. BroADen reported a 40% mean reduction in peak pruritus NRS but no itch responder rate, so BROADEN2 supplies the first KT-621 itch comparison.
Three limits on the concession. Route is one part of treatment burden, and each of these caps what an oral can charge for it.
- The oral comparator is a JAK. Upadacitinib 15 mg delivers a 36.9pp IGA delta in Measure Up, with a boxed warning and monitoring, which the same study values at ~21pp to avoid.
- Long-acting injectables narrow the convenience gap. Zumilokibart (APG777, now AbbVie’s) showed IGA 0/1 deltas of 17.6pp and 35.1pp in its Ph2 APEX trial (Appendix A5).
- Payers set the entry point. The 2026 Aetna Dupixent policy requires prior topical failure, so an oral STAT6 degrader likely enters through the same step logic.
Safety is not yet established. 0 of 22 patients had conjunctivitis in 28 days, which cannot exclude a true rate up to ~13% (Exhibit 7). Management expects conjunctivitis in the Ph2b (CEO, Citi summit, 10 Sep) and says dermatologists do not stop dupilumab for it, a second-hand claim.
Preference, uptake, payer access and share are four different steps, and the study measures only the first, so the Section 4 share inputs are placeholders. KT-621’s commercial value depends on how much efficacy and itch relief it preserves, not on being oral alone.
4. At $112.50, KYMR needs more than a good BROADEN2 result
A drug that works and a stock that works are different things. KT-621 could deliver a respectable dupilumab-like BROADEN2 result and still leave a valuation problem: at $112.50, KYMR trades ~10% above our $102 line-A value, so the market needs either a larger AD franchise than we underwrite or evidence that KT-621 earns more than line A.
The first is a commercial assumption and the second a clinical one. BROADEN2 can inform the second. It cannot settle the first.
The $6.9B question
Holding line-A inputs constant, $112.50 needs an unrisked AD peak of ~$6.9B, against $6.1B in our model. At our $32,000 net price that is a ~23% share of the treated pool rather than 20%, or ~80% approval from the Ph3 start rather than 70%. Neither is extraordinary alone: the stock needs a somewhat larger franchise than has been established, not an implausible one. After a line-B result (25pp to 30pp) our inputs support $4.6B and the price needs ~$8.2B, so a modest efficacy miss would need a much larger commercial outcome to offset it.
Dupixent is the yardstick. Sanofi reported €15.7B of FY2025 sales ($17.8B per Regeneron), and both companies describe 2026 as annualizing at more than €20B (~$23B at EUR/USD 1.15, our conversion), flattered by a favorable Q2 gross-to-net adjustment; neither guides 2026 sales. AD is not reported separately. At 45% to 55% of the franchise, our assumption, Dupixent AD was ~$8B to ~$10B in 2025 and ~$10B to ~$13B at the 2026 run-rate, and third-party estimates vary more widely. $6.9B is therefore not beyond precedent, but it is not small: Dupixent built its position through years of indication expansion, physician adoption and payer access, which convenience does not transfer automatically.
Consensus offers a bigger franchise, and a bigger extrapolation
Kymera consensus reaches $3.16B in 2033 and is still growing 70% Y/Y, too early to call a peak. We extend the series to ~2038, consistent with a 7 to 8 year ramp; everything after 2033 is a Clinaptis extrapolation, not consensus (Appendix A4), and the method matters. Fading the growth rate to zero gives ~$10.2B of revenue and, at 80% to 85% in AD, an AD peak of $8.2B to $8.7B. Fading the annual dollar increment gives $4.6B to $4.9B.
The first case is comparable to Dupixent’s 2025 AD sales and 65% to 85% of its 2026 run-rate, reached ~8 years after launch. It needs a 27% to 29% share of the treated pool against our 20% and carries line A to $127–133 and line B to $112–118; the restrained extension gives $85–88 and $76–79. The gap does not come from what analysts forecast through 2033. It comes from what we assume afterwards.
Management frames the market more broadly still: ~5% penetration of 6.5M to 9.5M moderate-to-severe US patients (Citi summit, 10 Sep). Our peak is ~12% to 14% of management’s 7M, so we assume less market expansion than management does. All comparisons use unrisked peaks, because consensus is not risk-adjusted.
What sits beneath the $112.50 price
The $112.50 reference price covers 107.98M fully diluted shares by the treasury method, an equity value of $12.15B. Against $1.646B of cash and securities at 30 Sep with no borrowings (3Q 2026 10-Q), EV is ~$10.50B (Appendix A2). Screens on basic shares show ~$9.6B of equity value and ~$8.0B of EV because they omit ~22.6M dilutive shares worth ~$2.5B at this price.
The cash is substantial but not surplus. Management guides to a runway into 2029 that funds “most” of the AD Ph3 and the start of the asthma Ph3 (Q2 2026 call, 5 Aug; reiterated in the 3Q release), so we expect a raise after data. Our per-share values hold the share count at 107.98M and deduct remaining AD development and launch spend.
Exhibit 6 breaks the value into its parts. Each readout line sets four inputs: approval probability from the Ph3 start, peak share of the US treated pool, net price, and a read-across multiplier on asthma and other type 2 indications. These are Clinaptis placeholders, not estimates from data (Appendix A2 has the full ledger, including line C’s inconsistent 90% Ph3-start assumption). The spread is wide: $102 a share for line A, $75 for B, $45 for C and $28 for D, with A+ at $155 as a reaction scenario rather than an efficacy category one Ph2b estimate can identify. AD is ~65% of line-A equity value and 76% of A+; the other programs add optionality but do not rescue a weak AD result.
Exhibit 6
Value per share runs from $28 on line D to $102 on line A and $155 in the A+ reaction scenario
Equity value per fully diluted share by BROADEN2 readout line on Clinaptis placeholders, with lines A and B re-run at the AD peak implied by extending consensus revenue to 2038 (a Clinaptis extension).
The AD franchise carries the spread: it is 76% of A+ equity and 65% of line A, while line D is over half cash ($15) with the rest residual value. A+ differs from A in share (30% against 20%), approval (80% against 70%) and read-across; it sits above the consensus-sized line A because consensus implies a similar ~27% to 29% share but at 70% approval. Totals are net of corporate cost. Values are conditional on Clinaptis placeholders, not a forecast.
View data
| Line | AD | Asthma | Other T2 | IRF5 | Platform | Corp. cost | Net cash | Equity, $M | $ / share | vs $112.50 |
|---|---|---|---|---|---|---|---|---|---|---|
| A+ | 12,780 | 455 | 566 | 394 | 1,500 | -554 | 1,646 | 16,785 | $155.44 | +38% |
| A | 7,180 | 387 | 492 | 394 | 1,500 | -554 | 1,646 | 11,044 | $102.28 | -9% |
| B | 4,381 | 342 | 443 | 394 | 1,500 | -554 | 1,646 | 8,150 | $75.47 | -33% |
| C | 1,389 | 206 | 295 | 394 | 1,500 | -554 | 1,646 | 4,875 | $45.14 | -60% |
| D | -134 | 70 | 148 | 394 | 1,500 | -554 | 1,646 | 3,069 | $28.42 | -75% |
Extended-consensus franchise (consensus revenue extended to 2038 by Clinaptis, AD at 80% to 85%): line A $127.03 to $133.34 and line B $112.25 to $117.65; conservative shape: A $84.69 to $88.13, B $75.96 to $78.91. Equity in $M; net cash is cash and marketable securities at 30 Sep 2026.
Outside AD, line A carries $2.8B (~$26 a share) before $0.55B of corporate cost: asthma (BREADTH Ph2b data in late 2027), other type 2 indications (management plans no further Ph2s in its 4 to 5 core indications but expects a different EoE path; Citi summit, 10 Sep, and Q2 2026 call), the IRF5 degrader KT-579 (lupus Ph1b planned by early 2027) and the platform and partnered programs (Sanofi’s KT-485 and Gilead’s KT-200). All four are placeholders: each $0.5B of platform value is $4.63 a share, and with none at all line A is ~$88.
Exclusivity is the other material assumption. Our base case runs to 2044, then post-expiry free cash flow falls 10% a year, adding ~$13 a share to line A; without it line A is ~$89, and expiry in 2040 or 2038 gives ~$95 or ~$91. We have not verified the KT-621 composition-of-matter expiry; the FY2024 10-K indicates STAT6 degrader patent families expire between 2038 and 2045 before any extensions. Both sensitivities are real, but the largest swing still comes from what KT-621 can earn in AD.
A+ is a commercial bet wearing a clinical label
A+ begins with a printed delta of 40pp or more, broadly comparable to upadacitinib 15 mg’s 39.5pp and 39.6pp in Measure Up 1 and 2. The endpoints differ, so this is a directional reference, not a matched comparison. The valuation asks a harder question: what if a 40pp print came to be treated as evidence of a durable, upadacitinib-like profile?
We lift approval from the Ph3 start from 70% to 80% (a 40pp Ph2b still leaves Ph3 efficacy, safety and regulatory risk), peak share from 20% to 30% at the same $32,000 net price, and the read-across multiplier to 1.15. The unrisked AD peak is then ~$9.1B, within our estimate of Dupixent’s 2025 AD sales, and equity value is $155.44 a share, 38% above the reference price.
Most of that step is share. Moving only share to 30% gives $138.57, only approval to 80% gives $112.65 and only the read-across to 1.15 gives $103.59, and share is the least anchored input in the note (25% gives $134.67, 35% gives $176.15). A 40pp print might change how investors see KT-621’s competitive position. It does not show that dermatologists will prescribe it to 30% of the treated pool, that payers will reimburse it favorably or that patients will stay on treatment. Those are separate commercial tests, and a Ph2b cannot run them.
The better dose can make the valuation look better than the drug
Even the 40pp threshold may mean less than it appears. Section 2 showed that at a true effect of 32.9pp, at least one of the two upper doses prints 40pp or more in 28% of trials under our central dispersion assumptions, and both do in only 7%. With the valuation stepping up $53 a share from A to A+, that gap matters.
We therefore value the possible readouts three ways, as conditional exercises on the Section 2 sampling probabilities, not probability-weighted estimates of efficacy. The reaction map pays each printed band its line value, including A+ whenever the better upper dose reaches 40pp. The second view withholds the A+ premium, so every print of 30pp or more earns line A. The third credits A+ only when both upper doses print 40pp or more.
| Valuation view, $ a share | Low | Central | High |
|---|---|---|---|
| Reaction map (A+ on the better dose) | 69.5 | 110.0 | 143.2 |
| No A+ premium | 68.1 | 95.2 | 101.8 |
| A+ only with two-dose confirmation | 68.2 | 99.0 | 126.4 |
Central dispersion: patient-level SD 38pp, 50 per arm, better of M and H, equal upper-dose effects and equal placebo SD. The Low, Central and High scenarios are not weighted against one another.
In Central, crediting A+ on the better dose adds ~$15 to the reaction map ($110.0 against $95.2), since only about a quarter of the 28% of trials with a 40pp best dose are both-dose results. In High, where the chance is 78%, the split adds $41; in Low it adds $1. The arithmetic does not create value. It shows how much value can be attributed to a noisy observation, and a stock can react as though a 40pp estimate establishes a 40pp drug. Our model should not make the same mistake.
What the reaction map does and does not tell us
The $110.0 Central value is not a fair value. It is the average value of the possible printed bands, each read as if it identified the drug. With an asymmetric payoff (B to A adds $27 a share, A to A+ another $53) that convention rewards favorable noise: a strong print lifts the average without KT-621 being any better. If the true effect is ~33pp, a holder who updates correctly on each print averages back toward line-A value, and the logic cuts the other way for noise-driven B and C prints.
The two-dose rule is not a formal posterior valuation. It is a practical evidence threshold: we will not award A+ economics because one arm lands in the upper tail of its sampling distribution. What would justify them is evidence beyond one number, set out in the scorecard (Exhibit 7), and then Ph3 design or payer evidence for a share above 25%. We assign no probabilities across Low, Central and High, so $69.5 to $143.2 is a range of conditional outcomes, not a fair value. The market may still trade the best number first and test its reliability later; that gap is the reason for showing three views.
Where we land
Cautious, not bearishAt $112.50 the stock sits between two commercial futures. On our line-A inputs it is worth $102, ~9% below the price. A franchise built on our growth-rate extension of consensus supports $127–133 for line A (13% to 19% above the price) and makes a line-B result roughly fair at $112–118; lines C and D are worth $45 and $28. Stressing one input at a time keeps line A between $84 and $120 (Appendix A2b), and a 15% to 25% peak share alone spans that range. The stock does not need an impossible outcome. It needs a combination of clinical result and commercial penetration that we cannot yet anchor.
A strong readout, a consistent dose response and convincing itch efficacy would improve the case for the franchise the price requires, but underwriting a share above 25% will need Ph3, prescribing and payer evidence. Weighting Low, Central and High equally gives illustrative averages of ~$88 to $108 across the three views, all below the price; those weights are not our probabilities and the averages are not a price target. For now, we are being asked to pay for some of that success in advance. Exhibit 7 sets out what would make us more willing to.
5. What would make us change our mind
A good BROADEN2 result is not by itself a reason to own KYMR at $112.50. We need evidence that changes the commercial assumptions in the price, not confirmation that STAT6 degradation works in AD.
The first test is a placebo-adjusted Wk16 mean EASI delta of 40pp or more in both upper doses. That would support the High scenario and give us reason to revisit A+ economics, though two point estimates are stronger evidence than one, not proof: at our Central true effect one dose reaches 40pp in 28% of trials and both in 7%. We read the adjusted delta, not the raw arm, because a higher placebo response shrinks the same drug-arm result.
The mean is only the start. A mean near 35pp with an EASI-75 delta near 30pp and a vIGA 0/1 delta of 20pp or less would describe less response depth than dupilumab’s: possibly useful for an oral, but not automatically a Dupixent-sized franchise. BroADen cannot settle this (EASI-75-to-mean ratio of 0.46 against 0.54 in SOLO and 0.66 in PROLEAD, on 21 uncontrolled patients), and absolute and ratio cut-offs disagree for 20 of 35 comparator arms (Appendix A3), so we apply no mechanical responder thresholds.
Three further reads decide how much we believe the mean. A strong EASI mean with weak itch relief would undermine the case for an oral, since patients value itch relief well above avoiding injections (Section 3). Safety is a denominator question: 0 of 22 ocular events bounds the true rate at 12.7%. And a strong high dose beside a much weaker second dose suggests investors are rewarding the better of two noisy estimates.
Exhibit 7 sets out the checks. The mean delta assigns the line, on Clinaptis categories rather than management targets; if a confidence interval spans two bands, the estimate has not distinguished them and we assign no single line.
Exhibit 7
A strong mean is the starting point; the second dose, response depth and itch determine how much we believe it
Clinaptis grading for the BROADEN2 readout. The mean-delta lines are Clinaptis judgments, not company bars; responder deltas are read against the comparator ranges in Appendix A3 and are not graded.
| Line | Mean delta, pp | Anchor |
|---|---|---|
| A+ | 40 or more | Upadacitinib 15 mg, Measure Up 1 and 2: 39.5 and 39.6; a reaction scenario that needs both doses to confirm; two point estimates above 40pp are stronger evidence than one, not proof |
| A | 30 to under 40 | Dupilumab Ph3, SOLO-1 and -2: 34.4 and 35.9; management expects data “in that ballpark of what dupilumab has shown at Week 16” (Q2 2026 call); no numeric target given |
| B | 25 to under 30 | Within ~10pp of dupilumab |
| C | 20 to under 25 | The JAK-without-the-safety-signal niche |
| D | under 20 | Below the stated bar |
| Read against | What to report |
|---|---|
| Dupilumab, lebrikizumab, upadacitinib, amlitelimab, rocatinlimab | Mean, EASI-75 and vIGA 0/1 deltas, each with its interval |
| KT-621, BROADEN2 | Report with intervals |
| Also read (not graded) | What to look for |
|---|---|
| Shape | EASI-75 and vIGA deltas divided by the mean delta. IL-4/IL-13: 0.96 to 1.08 and 0.67 to 0.78; JAK: 1.18 and 0.93; OX40: 0.95 and 0.53. A low ratio describes the shape of the response, not a commercial verdict, and the ratios are descriptive: SOLO means use available data while EASI-75 uses the full analysis set, so they mix populations. |
| Itch | Peak-pruritus responder delta (4-point NRS improvement), baseline NRS and the week of first separation. BroADen did not report an itch responder rate, so this is the first KT-621 itch comparison. |
| Safety | Counts, denominators and exposure windows, active minus placebo with intervals: ocular events, infections, discontinuations (AE-related and all-cause separately). BroADen’s 0 of 22 ocular events over 28 days bounds the true risk at 12.7% (one-sided 95%). |
| Dose response | Does the data pick a Ph3 dose, does the high dose do worse, does benefit-risk favor one. A plateau between M and H is not by itself a weakness; an isolated high-dose win beside a flat or worse second dose points to the noisy-best-arm reading, not to A+ economics. |
| Endpoints promised | Management says topline will include EASI-50, EASI-75, IGA 0/1, pruritus NRS and safety (Q2 2026 call, 5 Aug). BroADen’s EASI-50 was 76%. The itch measure and its format are not specified. |
| Not disclosed in advance | Baseline EASI, prior-biologic share and any split by it, regional mix, screen-failure rate and the open-label rollover rate. Management declined to give them (Q2 2026 call, 5 Aug), so read the delta beside whatever baseline the release shows. |
Report each delta with its confidence interval and the disclosed statistical success (analysis set, estimand and multiplicity procedure are not yet disclosed). If the confidence interval spans two bands, do not assign the result to a single band. Itch comparator rows are left out until they can be taken from primary papers.
View data
| Itch comparator (peak-pruritus NRS, 4-point responders) | Status |
|---|---|
| SOLO-1 / -2, Measure Up 2 | Values rest on secondary summaries; confirm definition in the papers |
| Measure Up 1, ADvocate-1 / -2 | Not yet retrieved from primary sources |
Even a convincing A+ result leaves a separate question: why should KT-621 take more than 25% of the treated AD pool? That needs Ph3 plans, durability and safety evidence and, eventually, prescribing and payer data, so we re-underwrite share separately from the efficacy band. We would not own KYMR for the readout alone.
Positioning into the readout (29 Oct 2026)
Short interest was 9.2M shares at the 15 Oct settlement, ~12% of float with ~13 days to cover (FINVIZ; MarketBeat shows ~13% and ~10 days), down from ~15% of float at 31 Mar though the position peaked at 10.5M shares on 15 Sep (Appendix A6), and the shares rose 42% on the BroADen data in Dec 2025. A line-A print could take the stock through our fundamental value on the day; we would treat that as a trade, not a re-rating.
Management plans to commercialize alone, considering a partner after the Ph2b only if it speeds or de-risks late-stage work (Stifel Immunology and Inflammation Forum, 24 Sep), with an end-of-Ph2 FDA meeting promptly after the AD data and first Ph3 dosing around mid-2027. The 3Q release (29 Oct) added positive KT-579 healthy-volunteer data (≥90% median blood IRF5 degradation), which says nothing about KT-621 efficacy.
Clinical versus commercialWe would become more constructive if both upper doses showed a durable-looking advantage, responder and itch data supported the mean, and development or access evidence made a larger franchise credible. A clinical win improves the case; the commercial case still has to be earned. BROADEN2 can tell us whether KT-621 deserves to compete with Dupixent. It cannot yet tell us whether KT-621 deserves to be valued like Dupixent.
Appendix
A1. Bridge and readout-noise assumptions
| Parameter | Range | Central | Basis |
|---|---|---|---|
| BroADen Day 29 mean EASI improvement | 63% | 63% | Company release, 8 Dec 2025 (reported) |
| Uncontrolled-to-blinded haircut | 0 to 25pp (Low 20, High 10) | 15pp | Judgment; no data identify it |
| Wk4 placebo improvement | 15% to 25% | 20% | Comparator trials (figure reads ±2pp) |
| Wk4 share of Wk16 effect (f) | 0.75 to 0.98 (Low 0.98, High 0.78) | 0.85 | Comparator trials: SOLO ~0.98, Ph2b ~0.78, CHRONOS ~0.66 |
| Placebo response at Wk16 | −30% to −45% | −35% | Judgment from 27-trial registry; sampling statements quoted at the −34% median of the six large Ph3s, drug-arm requirements at −35% |
| κ (drug arm moves with placebo) | 0, 0.5, 1 | none | Scenarios, unweighted |
| Dose structure | M/H 0.9 to 1.0; L/H 0 to 0.5 | 1.0; 0.25 | Management statement (Stifel forum, 24 Sep 2026) plus judgment |
| Patient SD, drug arm | 28 (SOLO-like), 35, 38, 42pp | 38pp | Lebrikizumab Ph2b Table 2 (250 mg arms 37.2 to 38.3pp; LS mean SDs, labelled LSSD in the FDA review, a proxy for patient-level dispersion); SOLO-1/2 registry Wk16, available-case (drug 26.3 to 27.8pp), likely understated |
| Placebo SD relative to drug arm | 1.0, 1.25, 1.5 | 1.0 | Lebrikizumab Ph2b placebo SD 56.5pp (LS mean SD, one trial); SOLO-1/2 placebo-to-drug ratio ~1.2 to 1.3 (33.7/26.3, 33.5/27.8pp) |
| Patients per arm | 50 | 50 | Design: ~200 patients, three doses and placebo; management says enrollment may run slightly over plan |
| Multiplicity | Holm, first step z=2.394 | Illustrative first-step threshold, stricter than nominal p<0.05; the SAP is undisclosed |
Reported inputs are company or trial disclosures; every other row is a Clinaptis judgment or derivation and is not a forecast.
Technical notes. At 21 evaluable patients the mean of the five BroADen patients below EASI-50 could lie anywhere from about −5% to just under 50%, so the aggregates cannot separate the response-shape hypothesis from open-label scoring, baseline or noise. The comparator f values are ~0.98 in the pooled SOLO trials, ~0.78 in the dupilumab and lebrikizumab Ph2bs and ~0.66 in CHRONOS (with topical steroids); PROLEAD’s match with SOLO is weak evidence that the haircut is small for an IL-4/IL-13 agent, but it is not a matched design. The lebrikizumab Ph2b used ANCOVA with multiple imputation and 44% placebo completion; at 55 and 60 patients per arm the standard error is ~7.2pp and ~6.9pp. BROADEN2 enrolls at 61 sites, including adolescents and sites in Japan and Korea (~20% of sites). Management’s dose comment (Stifel forum, 24 Sep) was that it chose a pharmacologically active middle dose, a higher dose “just to make sure” and a lower dose where it expects less activity; with the low dose at 25% of the high-dose effect, M and H sit at ~33pp and L at ~8pp. Management also said at least one recent AD trial posted a placebo response “in the 70s”, which we cannot verify. Prior sensitivity (Section 2): pooling two equal-effect doses gives a standard error of ~6.6pp against ~7.6pp for one contrast. With a normal prior centered at 33pp and prior SDs of 5pp to 9pp, two observed 40pp doses update to ~36pp to ~38pp and a 40pp and 30pp pair to ~34pp. These are illustrative sensitivities, not empirical priors.
Appendix A1, figure
The same 63% Ph1b mean supports a Wk16 effect of ~21pp to ~51pp at f=0.85, depending on the open-label haircut
True Wk16 placebo-adjusted effect for dose M by open-label haircut and Wk4 share f, at a Wk4 placebo response of 20%. Contours mark 20, 25, 30 and 40pp.
The grid shows sensitivity, not likelihood: the 63% BroADen Day 29 mean is open-label and uncontrolled, no data identify the haircut or f, and the three scenarios are unweighted.
View data
| Haircut, pp | f 0.70 | f 0.75 | f 0.78 | f 0.80 | f 0.85 | f 0.90 | f 0.95 | f 0.98 | f 1.00 |
|---|---|---|---|---|---|---|---|---|---|
| 0 | 61.4 | 57.3 | 55.1 | 53.8 | 50.6 | 47.8 | 45.3 | 43.9 | 43.0 |
| 5 | 54.3 | 50.7 | 48.7 | 47.5 | 44.7 | 42.2 | 40.0 | 38.8 | 38.0 |
| 10 | 47.1 | 44.0 | 42.3 | 41.2 | 38.8 | 36.7 | 34.7 | 33.7 | 33.0 |
| 15 | 40.0 | 37.3 | 35.9 | 35.0 | 32.9 | 31.1 | 29.5 | 28.6 | 28.0 |
| 20 | 32.9 | 30.7 | 29.5 | 28.8 | 27.1 | 25.6 | 24.2 | 23.5 | 23.0 |
| 25 | 25.7 | 24.0 | 23.1 | 22.5 | 21.2 | 20.0 | 18.9 | 18.4 | 18.0 |
A2. Line-to-value inputs and equity bridge
| Input or output | A+ (reaction scenario) | Line A | Line B | Line C | Line D |
|---|---|---|---|---|---|
| Observed delta band | ≥40pp | 30 to <40pp | 25 to <30pp | 20 to <25pp | <20pp |
| Ph3 start probability | 100% | 100% | 100% | 90% | 100% |
| Approval probability from Ph3 start | 80% | 70% | 60% | 45% | 25% |
| Peak share of treated pool | 30% | 20% | 15% | 10% | 5% |
| Net price per patient-year | $32,000 | $32,000 | $32,000 | $28,000 | $24,000 |
| Read-across multiplier (asthma, other type 2) | 1.15 | 1.0 | 0.9 | 0.6 | 0.3 |
| AD unrisked peak, WW ($M) | 9,117 | 6,078 | 4,558 | 2,659 | 1,140 |
| AD rNPV including terminal value ($M) | 12,780 | 7,180 | 4,381 | 1,389 | −134 |
| Asthma ($M) | 455 | 387 | 342 | 206 | 70 |
| Other type 2 indications ($M) | 566 | 492 | 443 | 295 | 148 |
| IRF5 degrader KT-579 ($M) | 394 | 394 | 394 | 394 | 394 |
| Platform and partnered programs ($M) | 1,500 | 1,500 | 1,500 | 1,500 | 1,500 |
| Corporate cost, present value ($M) | −554 | −554 | −554 | −554 | −554 |
| Net cash, 30 Sep 2026 ($M) | 1,646 | 1,646 | 1,646 | 1,646 | 1,646 |
| Equity value ($M) | 16,785 | 11,044 | 8,150 | 4,875 | 3,069 |
| Per fully diluted share (107.98M) | $155.44 | $102.28 | $75.47 | $45.14 | $28.42 |
| Against the $112.50 reference price | +38% | −9% | −33% | −60% | −75% |
Line C’s 90% Ph3 start probability is inconsistent with the 100% used in every other line (an input inconsistency, not a view); at 100% line C is worth $46.60 rather than $45.14, and a 40% Ph3 start probability in line D would add ~$0.6 a share.
Net cash is cash and marketable securities at 30 Sep 2026 of $1,645.9M ($373.4M cash and equivalents, $597.8M current and $674.7M non-current marketable securities; $6.2M of restricted cash excluded) with no borrowings (3Q 2026 10-Q). The model’s valuation date is ~27 Oct (82% of 2026 elapsed), so only October cash burn is unrecognized. Fully diluted shares of 107.98M at the $112.50 reference price are 85.37M basic (23 Oct), 15.82M pre-funded warrants, 8.63M options at a $44.30 weighted-average strike (5.23M net by the treasury method), 1.50M RSUs and 88K performance options at a $30.17 strike (0.06M net).
Other Base-column inputs: 4.5M diagnosed US AD patients growing 2% a year; 40% moderate-to-severe and systemic-eligible; advanced-systemic penetration 25% in 2026 rising to 45% over 10 years; 0.8 paid patient-years per patient; 3% annual net-price erosion; ex-US revenue at 0.4 times US; gross margin 92%, S&M 25% and G&A 5% of sales; tax 18%; discount rate 11%; loss of exclusivity 2044 (2043 Bear, 2045 Bull) and free cash flow falling 10% a year afterwards (20% Bear, 5% Bull). Other type 2 indications: 15% PoS on $2.5B peak. IRF5: 10% PoS on $3.0B peak. All Clinaptis placeholders. The market inputs give ~450K treated patients in 2026 against management’s ~400K dosed with an advanced systemic therapy, which management states as ~5% of 6.5M to 9.5M moderate-to-severe US patients (Citi summit, 10 Sep); our pool is the 1.8M systemic-eligible subset.
Partnered programs. The $1.5B platform and partnered placeholder is a Clinaptis judgment. The terms below show what is disclosed and do not derive it.
| Program | Terms | Source |
|---|---|---|
| Sanofi, KT-485 (IRAK4) | Up to $955M of further development and commercial milestones at 30 Jun 2026 (the 9 Jun release says up to $975M, which reconciles arithmetically with the $20M milestone earned in Jun; we have not confirmed that in the filings). Tiered royalties from high single digits to high teens. Option to fund 50% of US development costs and then share US profits equally. $20M milestone earned on first dosing (Jun 2026), recognized in Q2 revenue and received in Q3. | 10-K FY2025, 10-Q Q2 2026 |
| Gilead, KT-200 (CDK2 glue) | $40M upfront (Jun 2025) and $45M option exercise (Apr 2026, received in Q2). Up to $665M of further milestones. Royalties from high single digits to mid-teens. Gilead holds global rights. | 10-K FY2025, release 9 Apr 2026 |
| All collaborations | No royalties received to date. | 10-K FY2025 |
A2b. Valuation stress, one input at a time
| Line | Input changed | Value tested | $ / share | vs line base |
|---|---|---|---|---|
| A | Peak share | 15% | 84.13 | −18.15 |
| A | Peak share | 17.5% | 93.20 | −9.07 |
| A | Peak share | 25% | 120.43 | +18.15 |
| A | Approval from Ph3 start | 60% | 91.91 | −10.37 |
| A | Approval from Ph3 start | 80% | 112.65 | +10.37 |
| A | Post-LOE value | off | 89.47 | −12.80 |
| A | LoE year | 2040 | 95.49 | −6.79 |
| A | LoE year | 2038 | 90.72 | −11.56 |
| A | Platform and partnered | $1.0B | 97.65 | −4.63 |
| A | Platform and partnered | $0.5B | 93.02 | −9.26 |
| A | Platform and partnered | none | 88.39 | −13.89 |
| A | Discount rate | 12% | 93.52 | −8.75 |
| A | Combined: mild downside | 17.5% share, $1.0B platform | 88.57 | −13.70 |
| A | Combined: downside | 15% share, 60% approval, no post-LOE, $0.5B platform | 58.47 | −43.80 |
| A | Combined: upside | 25% share, 75% approval | 126.91 | +24.63 |
| B | Peak share | 20% | 91.03 | +15.56 |
| B | Peak share | 25% | 106.59 | +31.11 |
| B | Approval from Ph3 start | 70% | 83.25 | +7.78 |
| B | Post-LOE value | off | 66.97 | −8.51 |
| B | Platform and partnered | $0.5B | 66.21 | −9.26 |
| B | Platform and partnered | none | 61.58 | −13.89 |
| A+ | Share only | 30% | 138.57 | −16.87 |
| A+ | Approval only | 80% | 112.65 | −42.80 |
| A+ | Read-across only | 1.15 | 103.59 | −51.86 |
| A+ | Half-step | 25% share, 75% approval, read-across 1.0 | 126.91 | −28.54 |
| A+ | Post-LOE value | off | 134.10 | −21.34 |
| A+ | Platform and partnered | $0.5B | 146.18 | −9.26 |
Line bases are $102.28 (A), $75.47 (B) and $155.44 (A+). Each row changes the named input (combined rows say what changes) in the Clinaptis Analytical Model and holds the rest at the line’s inputs; a full recalculation reproduces every published line value. To reach $112.50, line A needs a ~23% peak share or a ~80% approval probability, and line B a ~27% share. Values are conditional on Clinaptis placeholders, not a forecast.
A2c. Reaction-map valuation by placebo dispersion
| Scenario | Placebo SD | D | C | B | A | A+ | Reaction map, $ | A+ withheld, $ | A+ if both doses ≥40pp, $ |
|---|---|---|---|---|---|---|---|---|---|
| Low | 1.0x | 17.5% | 24.1% | 27.7% | 27.9% | 2.7% | 69.5 | 68.1 | 68.2 |
| Central | 1.0x | 1.0% | 4.6% | 13.9% | 52.4% | 28.0% | 110.0 | 95.2 | 99.0 |
| High | 1.0x | 0.0% | 0.1% | 1.2% | 20.9% | 77.8% | 143.2 | 101.8 | 126.4 |
| Low | 1.5x | 24.0% | 19.6% | 21.3% | 27.9% | 7.2% | 71.5 | 67.6 | 68.5 |
| Central | 1.5x | 4.1% | 7.5% | 14.2% | 41.1% | 33.0% | 108.7 | 91.1 | 98.4 |
| High | 1.5x | 0.3% | 1.0% | 3.4% | 23.4% | 71.9% | 138.8 | 100.6 | 125.5 |
| Low | SOLO-like (28/34pp) | 16.3% | 29.0% | 31.7% | 22.2% | 0.7% | 65.6 | 65.2 | 65.2 |
| Central | SOLO-like (28/34pp) | 0.4% | 3.5% | 14.7% | 61.0% | 20.3% | 106.8 | 96.0 | 98.9 |
| High | SOLO-like (28/34pp) | 0.0% | 0.0% | 0.6% | 21.2% | 78.2% | 143.7 | 102.1 | 129.1 |
Moved from the former Exhibit 1 data table. Reaction map: each printed band valued at its line value (A+ at $155.44). A+ withheld: every print of 30pp or more valued at line A. A+ if both doses ≥40pp: A+ credited only when both upper doses print 40pp or more. D to A+ are the sampling odds of each printed band for the better of two doses. Placebo SD 1.0x and 1.5x are multiples of the 38pp drug-arm SD; SOLO-like is a 28pp drug arm and 34pp placebo arm. Percentages are sampling odds for one true effect, not Clinaptis forecasts of the outcome, and we assign no probabilities across scenarios. Values are Clinaptis estimates on stated placeholders (modeled, not observed data).
A3. Comparator responder deltas
| Comparator, monotherapy, Wk16 | Mean delta, pp | EASI-75 delta, pp | vIGA 0/1 delta, pp |
|---|---|---|---|
| Dupilumab 300 mg Q2W, SOLO-1 / SOLO-2 | 34.4 / 35.9 | 36.6 / 32.3 | 27.6 / 27.6 |
| Lebrikizumab 250 mg Q2W, ADvocate-1 / -2 | 38.3 / 33.5 | 42.6 / 34.0 | 30.4 / 22.4 |
| Upadacitinib 15 mg, Measure Up 1 / 2 | 39.5 / 39.6 | 53.3 / 46.8 | 39.7 / 34.1 |
| Amlitelimab Ph2b, four doses | 22.2 to 32.1 | 27.1 to 31.5 | 9.0 to 20.2 |
| Rocatinlimab Ph2b, four doses | 33.3 to 46.1 | 28.4 to 43.3 | 13.6 to 29.0 |
Comparator deltas are placebo-adjusted and not head-to-head. Means and responder rates use different analysis sets and missing-data rules in several trials (SOLO mean deltas use available data, while EASI-75 and vIGA use the full analysis set with rescue and missing counted as non-response). Source: SOLO-1/2 pooled analysis, ADvocate-1/2, Measure Up 1/2, and the amlitelimab and rocatinlimab Ph2b publications and registry results.
A4. Extending consensus revenue to 2038
| Total revenue, $M | 2033 | 2034 | 2035 | 2036 | 2037 | 2038 |
|---|---|---|---|---|---|---|
| Consensus (average) | 3,159 | |||||
| Shape 1: growth rate fades linearly to 0 (Clinaptis extension) | 3,159 | 4,924 | 6,987 | 8,939 | 10,188 | 10,188 |
| Shape 2: growth rate ×0.7 a year (Clinaptis extension) | 3,159 | 4,703 | 6,313 | 7,825 | 9,137 | 10,210 |
| Shape 3: annual increment fades linearly to 0 (Clinaptis extension) | 3,159 | 4,198 | 4,978 | 5,497 | 5,757 | 5,757 |
| At 80% to 85% of total in AD | Shape 1 | Shape 2 | Shape 3 |
|---|---|---|---|
| AD unrisked peak ($M) | 8,150 to 8,660 | 8,168 to 8,678 | 4,606 to 4,893 |
| Equivalent peak share at $32,000 net price | 27% to 29% | 27% to 29% | 15% to 16% |
| Line A, $ per share | 127.03 to 133.11 | 127.24 to 133.34 | 84.69 to 88.13 |
| Line B, $ per share | 112.25 to 117.46 | 112.43 to 117.65 | 75.96 to 78.91 |
Consensus revenue is the average annual figure supplied for 2027 to 2033 ($39M, $37M, $67M, $310M, $940M, $1,860M, $3,159M). Years after 2033 are a Clinaptis extrapolation, not consensus.
A5. Emerging atopic dermatitis candidates
| Candidate | Sponsor | Mechanism, route, dosing | Stage, Oct 2026 | Efficacy as reported by sponsor or journal | Read |
|---|---|---|---|---|---|
| Injectable biologics with efficacy data | |||||
| Zumilokibart (APG777) | AbbVie (bought Apogee, closed 3 Sep 2026, $10.9B equity) | Half-life-extended anti-IL-13 mAb, SC; maintenance every 12 or 24 weeks | Ph3 start 2H 2026 | APEX Part B Wk16, mid dose (planned Ph3 dose, n=85) vs placebo (n=88): EASI-75 65.9% vs 23.4% (Δ41.9pp, sponsor-reported placebo-adjusted; the raw rates subtract to 42.5pp); vIGA 0/1 46.0% vs 10.9% (Δ35.1pp); conjunctivitis 10.6% (all terms). Part A: EASI-75 66.9% vs 24.6%. | Closest long-acting rival; conjunctivitis not avoided |
| Tilrekimig (PF-07275315) | Pfizer | IL-4/IL-13/TSLP trispecific, SC; Q2W (Stage 1), Q4W (Stage 2) | Ph3 dosing (3 studies, one vs dupilumab) | Biologic-naive adults, Wk16 EASI-75. Stage 1, 450 mg Q2W (n=41 vs 20): 62.5% vs 19.9%. Stage 2, Q4W: 58.5% (400 mg), 61.0% (200 mg), 47.8% (50 mg) vs 9.1%; Δ49.4, 51.9, 38.7pp. | Small arms; mean EASI not in the release |
| Ompekimig (PF-07264660) | Pfizer | IL-4/IL-13/IL-33 trispecific, SC | Ph2 | Stage 1 high dose met the primary endpoint (EASI-75 vs placebo); figures not seen. | No numbers to compare |
| Rademikibart (CBP-201) | Connect Biopharma; Simcere (China Ph3) | IL-4Rα mAb, SC; Q2W or Q4W (Q4W tested in Ph2) | WW001 Ph2 (2020 to 2021); China Ph3 reported Mar 2026 | WW001 (n=226) Wk16 mean EASI change: −63.0% (300 mg Q2W), −63.5% (300 mg Q4W), −57.6% (150 mg Q2W) vs placebo −39.7%; Δ23.3, 23.8, 17.9pp (our arithmetic). Conjunctivitis 2.9% vs 0%. China Ph3 (n=259, Q2W): EASI-75 74.2% vs 34.4% (Δ39.8pp); IGA 0/1 47.7% vs 17.6% (Δ30.1pp). | Ph2 Δ~23pp is line C on our scale; the pitch is Q4W, not depth |
| Lunsekimig (SAR443765) | Sanofi | IL-13/TSLP bispecific nanobody, SC; Q2W or Q4W | VELVET Ph2b done; AD exploratory, respiratory Ph3 | VELVET Wk24 (n=150; ~37 per arm, 39 pooled placebo): 300 mg Q4W mean EASI −67.7% vs placebo −43.7% (Δ23.9pp, adjusted p=0.036 vs 0.025 needed); Q2W Δ18.2pp; 600 mg Q4W Δ10.1pp. EASI-75 62.2%, 58.3%, 39.5% vs 17.9%. Primary missed. | Sanofi cites placebo response; 95% CI on Δ runs 3 to 45pp |
| Amlitelimab | Sanofi | OX40L mAb, SC; Q4W or Q12W | Discontinued in AD, 24 Jul 2026 | COAST 1 Wk16 EASI-75: 35.9% (Q4W), 39.1% (Q12W) vs 19.1%. Sanofi: the evidence “does not support further development” in AD. | OX40 shape reference in Exhibit 7 |
| Oral STAT6, KT-621’s class: no efficacy data | |||||
| NX-3911 (SAR448272) | Nurix; Sanofi | Oral STAT6 degrader | Sanofi-partnered; Ph1 dosing started (Kymera call, 5 Aug) | No clinical AD efficacy found in sources reviewed. | Closest mechanism |
| REX-8756 (SAR448755) | Recludix; Sanofi | Oral STAT6 inhibitor | Ph1 healthy volunteers; data 2H 2026 | Mouse AD model: efficacy “comparable” to anti-IL-4/IL-13 antibodies (company, preclinical). | |
| PF-08049820 | Pfizer | STAT6 inhibitor | Ph2 in AD (NCT07216027), n=165, recruiting; primary completion Feb 2028 | None disclosed. | |
Every figure is sponsor- or journal-reported, from different trials with different populations (biologic-naive vs unselected), timepoints (Wk16 unless stated), imputation rules and placebo rates, so none is comparable with the others or with BROADEN2. Most sponsors report EASI-75, not the mean EASI delta our lines use. In IL-4/IL-13 antibodies the EASI-75 delta runs about 1.0 times the mean delta (Exhibit 7), but VELVET (EASI-75 delta 44.5pp as reported, 44.3pp from the quoted rates, against a mean delta of 23.9pp, ~37 patients per arm) shows the ratio is unstable in small arms. Omitted because they do not bear on the share or endpoint case: the oral non-STAT6 programs orismilast (ADESOS Ph2b, primary missed) and EP262 (EASE Ph2a, no signal) and the preclinical STAT6 program EPS-3903. Not tabulated: Chinese IL-4Rα programs beyond rademikibart (for example MG-K10, Ph3), eblasakimab, rocatinlimab (Ph3; Ph2b in Exhibit 7) and earlier-stage candidates.
A6. Share price and short interest
Appendix A6
KYMR has risen 57% this year against 29% for the XBI, and short interest has eased to ~12% of float
Daily closes indexed to 100 at 31 Dec 2025. KYMR leads the XBI by 28 points and the Nasdaq Biotechnology Index (NBI) by 30. Shares sold short ranged from 8.7M to 10.5M this year, peaking on 15 Sep, and were 9.2M on 15 Oct.
View data
| Month-end close | KYMR, $ | KYMR index | XBI index | NBI index |
|---|---|---|---|---|
| 31 Dec | 77.81 | 100.0 | 100.0 | 100.0 |
| 30 Jan | 72.69 | 93.4 | 102.3 | 102.5 |
| 27 Feb | 91.35 | 117.4 | 104.5 | 105.9 |
| 31 Mar | 83.29 | 107.0 | 104.8 | 102.3 |
| 30 Apr | 81.07 | 104.2 | 107.7 | 102.7 |
| 29 May | 81.42 | 104.6 | 112.1 | 105.0 |
| 30 Jun | 114.67 | 147.4 | 129.8 | 115.0 |
| 31 Jul | 103.40 | 132.9 | 120.6 | 113.1 |
| 31 Aug | 118.81 | 152.7 | 133.3 | 124.8 |
| 30 Sep | 111.79 | 143.7 | 129.3 | 124.7 |
| 28 Oct | 122.51 | 157.4 | 129.2 | 127.1 |
A7. Sources
Company releases and filings (design, timing, cash, shares, KT-579, partnered terms). BroADen Ph1b (8 Dec 2025); BROADEN2 enrollment and topline timing (25 Jun 2026); Q2 2026 results (5 Aug 2026); 3Q 2026 10-Q and release (29 Oct 2026; cash and securities at 30 Sep, share count at 23 Oct, last-patient visit, KT-579 healthy-volunteer data); AAD 2026 late-breaking presentation; Q2 2026 10-Q; FY2025 10-K; FY2024 10-K (patents); Kymera 1Q 2026 release (30 Apr 2026).
Management commentary. Kymera Q2 2026 earnings call (5 Aug 2026), Citi Back to School Summit (10 Sep 2026) and Stifel Immunology and Inflammation Forum (24 Sep 2026) transcripts (user-supplied transcription).
Placebo registry, dispersion and bridge (Section 1, Section 2, Exhibits 2 to 4, A1). Placebo registry and SOLO-1/2 Wk16 SDs and analysis sets from ClinicalTrials.gov results (NCT numbers in the Exhibit 2 data table); SOLO-1/2 pooled analysis; lebrikizumab Ph2b (Guttman-Yassky, JAMA Dermatology 2020, Table 2); PROLEAD (Dermatology and Therapy 2023); Clinaptis analytics files Analytics/d1 (T2 bridge grid, T4 placebo registry, T5 and T6 sampling).
Comparator deltas (Exhibit 7, A3). Measure Up 1 and 2; ADvocate-1/2; SOLO-1/2 pooled analysis as above; amlitelimab and rocatinlimab Ph2b publications and registry results.
Preference and access (Section 3, Exhibit 5). Thomas et al., BMJ Open 2022;12:e058799; Aetna 2026 Dupixent policy.
Valuation (Section 4, Exhibit 6, A2, A2b, A4). Sanofi Q4/FY2025 (29 Jan 2026) and Q2 2026 (30 Jul 2026) releases; sell-side consensus revenue (Fintel average); Clinaptis Analytical Model.
Price and positioning (A6). FINVIZ short interest (15 Oct 2026 settlement, published 26 Oct) and Yahoo Finance daily closes.
Emerging candidates: Apogee APEX Part B deck (26 May 2026); AbbVie completes Apogee acquisition (3 Sep 2026); Pfizer tilrekimig Ph2 (1 Oct 2026) and Stage 1 release (9 Mar 2026); rademikibart WW001 (JACI 2024); rademikibart China Ph3 (30 Mar 2026); Sanofi VELVET poster (Sep 2026); amlitelimab discontinuation (24 Jul 2026) and COAST 1 (28 Mar 2026); Recludix REX-8756; Pfizer Q4 2025 pipeline update.