NATiV3, Inventiva’s Phase 3 in MASH, can meet its primary endpoint and still leave Inventiva worth €2.62 a share against a €3.36 price, if the win is narrow: a composite gain under 10 points or a weak fibrosis result. We put a primary hit at 78%. On our branch values the price breaks even at 63% to 77%, depending on what a persuasive win is worth.

NATiV3 in four numbers

A hit is not enough. The stock needs a persuasive result.

Primary endpoint met

78%

Our odds. The price breaks even at about 68% on our branch values.

Modeled composite gain

+12pts

1,200 mg, 72 weeks; fibrosis +13. About 5 to 6 clears the statistical bar.

Value if the win is narrow

€2.62

A persuasive result is worth €4.24; an ESSENCE-level one €6.31.

Probability-weighted return

+1% to +16%

About one outcome in five loses more than half the share price.

Neutral, upside skew

Break-even needs the persuasive branch worth €0.95B of enterprise value, about where Inventiva trades today, against the €1.6B we assume.

Modeled, not observed data. Price €3.36 at the 10 November close. Values per share are after royalty certificates. Return range spans a persuasive-branch enterprise value of €1.0–2.0B. The 68% is a break-even probability on our values, not a market-implied figure. Source: Clinaptis NATiV3 forecast and valuation models.

Lanifibranor came out of the Fournier research site at Daix that Inventiva’s founders, then running research for Abbott, took over in 2012. After the 2025 decision to stop all other research, it is the asset on which almost the whole company now depends.

Inventiva (IVA; €3.36, ~€1.05B on 313.7M shares including penny warrants) is weeks from the readout of NATiV3, a Phase 3 in MASH (metabolic dysfunction-associated steatohepatitis, the fatty-liver disease with inflammation and scarring formerly called NASH). The trial randomised 1,009 patients with F2–F3 fibrosis (moderate to advanced scarring, short of cirrhosis) to 72 weeks of placebo or lanifibranor at 800 mg or 1,200 mg; a separate 410-patient exploratory cohort, mostly F1 and F4, sits outside the efficacy analysis. Its primary endpoint requires MASH resolution and at least one stage of fibrosis improvement in the same patient. That is stricter than FDA requires, since FDA accepts either, and we infer it was chosen to meet the European regulator’s requirement that both be shown. Management said on 9 November that data should arrive within about eight weeks.

A primary hit is the lowest of three bars. The prespecified NATIVE result was a composite gain of roughly 24 points at 24 weeks. We model a 12-point gain at 72 weeks (80% range 0 to 27), with fibrosis at 13, enough to clear the statistical bar of about 5 to 6 points on our assumptions; the primary endpoint is met in about 78% of simulated trials. A persuasive win needs a composite gain of at least 10 points, significance on both 1,200 mg key secondaries (resolution and fibrosis) and a fibrosis gain of at least 10. About 52% of simulated trials get there, including those that clear the next bar. Matching semaglutide’s ESSENCE result, a composite gain near 17 with fibrosis near 15, is the competitive histology bar, at roughly 30%. Weight gain and tolerability then decide whether a result supports an oral add-on to GLP-1s. Exhibit 1 sets out each bar and what it is worth per share.

A narrow hit is worth less than the price. The probability-weighted trade breaks even at €3.36 if the persuasive-but-undifferentiated branch is worth about €0.95B of enterprise value, roughly where Inventiva trades today, against the €1.6B we assume. After a modeled deduction for the royalty certificates, the expected return runs from +1% to +16% as that value moves from €1.0B to €2.0B. About one outcome in five loses more than half the share price; about half gain a quarter or more. The financing sharpens the tails: a secondary-only result can make the warrants exercisable while putting the debt in default, and cash then depends on holders choosing to exercise. We stay Neutral, with the skew to the upside.

We would be less surprised by an ESSENCE-level result (about 30%) than by a primary miss (about 22%, secondary-only results included).

Clinaptis house view, NATiV3 forecast (modeled, not observed data)

Exhibit 1

NATiV3 has three bars to clear; our central case clears two

Each bar tested at our central week-72 forecast: 1,200 mg composite 23% against 11% on placebo, a 12-point gain, with a 13-point fibrosis gain. Probabilities are model outputs, not trial facts. Value per share applies if that is the highest bar cleared.

Central case is the median of the Clinaptis forecast (40,000 simulated trials, about 336 patients an arm, missing biopsies as failures). Probabilities are cumulative. About 5 points clears p<0.05 at an 11% placebo rate, and about 6 if 1,200 mg is the only dose to win (Hochberg then tests it at 0.025). Values per share are the Exhibit 6 branch values.

Source: Clinaptis NATiV3 forecast and valuation models (modeled, not observed data); trial publications and Inventiva filings (see Sources).

01NATIVE

NATIVE: the case rests on 1,200 mg

NATIVE makes its case at 1,200 mg. The 800 mg arm is the weak point, and the published headline raises the estimated gain by 1.9 points.

At 1,200 mg, 31.3% of patients met the composite at week 24 on the trial’s prespecified method (MASH resolution plus at least one stage of fibrosis improvement), against 7.4% on placebo: a gain of 23.9 points (95% CI +12 to +35). Both components moved, resolution by 26.1 points and fibrosis improvement by 18.7, or 17.8 in the F2–F3 patients NATiV3 enrolled. The composite was a secondary endpoint, so its p-values are unadjusted for multiple testing. We rely on the consistency across endpoints, and on 1,200 mg meeting the one endpoint the trial did protect (a two-point fall in the biopsy activity score).

The weak spot is 800 mg. It missed that primary activity endpoint (p=0.06 on the plan’s method, 0.07 in the paper) and improved fibrosis by only 4.3 points over placebo (95% CI −9 to +17), or 2.5 points in the F2–F3 patients. Its 13.1-point composite gain came almost entirely from MASH resolution.

The headline method raises the estimated gain by 1.9 points. The NEJM paper reports composite response of 9%, 25% and 35% on placebo, 800 mg and 1,200 mg, using multiple imputation for missing biopsies. The protocol and statistical plan, dated 13 May 2020, a month before topline, named non-responder imputation (a missing biopsy counts as a failure) as the primary method. The supplement labels the imputed analysis “post hoc”, yet the paper calls it “the primary analyses that we report here”. On the plan’s method the 1,200 mg gain is 23.9 points against 25.8 in the headline, which does not move the forecast. The prespecified 1,200 mg result is the signal we carry forward.

Exhibit 2

NATIVE: 1,200 mg moved all three endpoints; 800 mg’s fibrosis gain was small and uncertain

Share of patients responding at week 24, counting missing biopsies as failures (the trial’s prespecified method). At 1,200 mg, seven in ten patients whose MASH resolved also improved fibrosis; 800 mg’s composite gain tracked its resolution gain, not fibrosis.

View data
Endpoint and methodPlacebo800 mg1,200 mg
Composite, non-responder (prespecified)7.4% (6/81)20.5% (17/83)31.3% (26/83)
Composite, multiple imputation (NEJM headline, post hoc)9.0%24.9%34.8%
Composite, complete cases9.0%23.6%35.1%
MASH resolution, non-responder18.5%32.5%44.6%
Fibrosis improvement, non-responder23.5%27.7%42.2%
Fibrosis improvement, F2–F3 subgroup29.8%32.4%47.6%

247 patients randomised (81 / 83 / 83); 24% had F0–F1 fibrosis. Non-responder imputation is the prespecified method; the paper’s headline multiple-imputation figures (9 / 25 / 35% composite) were post hoc. Only the primary activity endpoint was multiplicity-controlled, so these gains are nominal. Gains are computed from unrounded rates.

Source: Francque et al., NEJM 2021 (Supplementary Appendix, protocol and SAP); Clinaptis calculations (see Sources).

02Translation

How much of the effect survives to 72 weeks

NATiV3 needs about 5–6 points of composite gain at 1,200 mg to clear our modeled statistical bar. We forecast 12, with fibrosis at 13.

That forecast has an 80% range of 0 to 27 points, against the 24 points NATIVE showed at 24 weeks, and the primary is met in 78% of simulations. NATiV3 can give back half of NATIVE’s gain and still clear the bar, but about one simulated trial in ten has no true benefit, so a hit is likely but not assured. The bar assumes about 336 patients an arm and counts missing biopsies as failures. The statistical analysis plan is unpublished. What could erase the margin is how much of NATIVE’s effect translates to Phase 3.

Five MASH programmes have a randomised Phase 2 and a Phase 3 on a matching endpoint. Semaglutide, resmetirom and obeticholic acid kept roughly three-quarters or more of their Phase 2 effect on the log-odds scale (Appendix A2). Elafibranor kept less and failed, and cenicriviroc reversed. We give those two failures half weight. That is a judgment: weighting them fully takes the hit rate to 70% and the expected return to about zero (A3), so readers should hold both numbers. NATIVE’s 1,200 mg signal was consistent across endpoints and met its multiplicity-protected activity primary, although the composite itself was nominal. Elafibranor’s Phase 2 signal rested on a post hoc definition and cenicriviroc’s on fibrosis alone. The weighting moves the hit rate more than any other setting we tested. The composite is built from the resolution and fibrosis effects and needs both in the same patient, so it shrinks more than either component. The translation factor bundles more than time: differences in population, endpoint definition and pathology readers between Phase 2 and Phase 3 all sit inside it. NATIVE is also a single-reader trial.

Fibrosis gets more time than the calibration allows for. Four of the five calibration pairs compare trials of similar length, so the factor carries no allowance for NATiV3 running three times longer than NATIVE. Time does not help both endpoints equally. Placebo fibrosis response barely moves with duration (23.5% at 24 weeks in NATIVE, about 19% to 22% at 72 weeks in ESSENCE, 20% and 19% at 24 and 96 weeks in HARMONY), while placebo resolution climbs (18.5% in NATIVE against 29% to 34% in ESSENCE). Active fibrosis effects can keep building: efruxifermin’s fibrosis gain rose from 21 points at week 24 to 30 at week 96 on an intention-to-treat basis. NATIVE’s 18.7-point fibrosis gain at only 24 weeks points to a direct antifibrotic effect that could do the same (inference). We therefore scale the fibrosis translation up by a quarter. That lifts the central fibrosis gain from 10 to 13 points and the chance of an ESSENCE-level result from about 21% to 30%.

Missing biopsies pull the other way. A missing biopsy counts as a failure, and our anchors embed roughly 15% missing (ESSENCE). Management guides dropout “comfortably below 30%”, so we assume 22%. That shrinks every arm’s response rate by about 8% and the composite gain by about a point. On the composite the fibrosis allowance and the dropout haircut roughly cancel, and the central gain stays at 12.

The hit rate depends most on how the two failed programmes are weighted

Probability that NATiV3 meets its primary endpoint under alternative translation settings. Our house view is the second column.

All five programmes, equal weight

70%

Both failures at full weight

House view

78%

Failures at half weight

Positive programmes only

90%

Failures excluded

Missing biopsies 15% / 28%

79% / 76%

Against our 22% assumption

Modeled, not observed data. 40,000 simulated trials per setting; full table, including the fibrosis allowance and erosion settings, in Appendix A3.

Placebo moves the hit rate far less. We centre the composite placebo rate at 11% (80% range 6% to 18%) on the missing-biopsies-as-failures basis, near the 13.6% ESSENCE ran on the same basis (our conversion from published counts). Fixing placebo near that level leaves the hit rate at 78%. GLP-1 use is a separate matter. ESSENCE excluded patients on a GLP-1 at entry, while 14% of NATiV3 patients were on a stable GLP-1 at baseline. ESSENCE’s analysis that set aside patients who started a GLP-1 during the trial left placebo resolution unchanged (34.2% against 34.3%). That addresses starts after randomisation, not baseline use, and we do not model the baseline difference.

At 800 mg we model a 4-point composite gain (80% range −1 to +12), nominally significant in about 40% of simulations. At 1,200 mg, fibrosis is the softer key secondary, with a gain of about 13 points that reaches significance in about 75% of simulations, against about 80% for resolution. The persuasive branch needs both to be significant, plus composite and fibrosis gains of at least 10 points, which is why it and the branch above sit near 52% when the primary sits near 78%. A primary hit says some of NATIVE’s effect survived. How much fibrosis benefit survives decides whether the result matters against semaglutide, and how large that gain has to be is the question for §3.

Modeled forecast

Exhibit 3

1,200 mg likely clears the composite; fibrosis and 800 mg are softer

Modeled week-72 response by endpoint and dose, non-responder basis. Central case: 1,200 mg composite 23% against 11% on placebo, a 12-point gain; the primary endpoint is met in about 78% of simulated trials.

1,200 mg800 mg
EndpointResponse, placebo to 1,200 mgGain, 80% rangeChance p<0.05GainChance p<0.05
Composite (primary)11% → 23%+12 (0 to +27)80%+440%
MASH resolution23% → 41%+17 (0 to +34)80%+960%
Fibrosis improvement19% → 33%+13 (0 to +29)75%+230%

Chance the observed 1,200 mg gain reaches each benchmark

EndpointLower barESSENCE bar
Composite≥10 pts persuasive floor 60%≥17 pts 35%
Fibrosis≥11.8 pts MAESTRO 100 mg 55%≥15.4 pts 40%
View data
EndpointPlacebo (80% range)800 mg1,200 mgGain, 800 mg (80% range)Gain, 1,200 mg (80% range)P(p<0.05) 800 / 1,200
Composite (primary)11% (6–18)15%23%+4 (−1 to +12)+12 (0 to +27)40% / 80%
MASH resolution, no fibrosis worsening23% (14–36)33%41%+9 (0 to +22)+17 (0 to +34)60% / 80%
Fibrosis ≥1 stage, no MASH worsening19% (11–31)22%33%+2 (−3 to +12)+13 (0 to +29)30% / 75%

Modeled, not observed data. Rates are medians; gains are placebo-adjusted. Probabilities are rounded to 5 points: green is 75% or more, yellow 40–74%, red below 40%. Assumes about 336 per arm, Hochberg across doses, missing biopsies (22%) as non-response and a fibrosis allowance for the longer trial (§2); the SAP is not public (A1).

Source: Clinaptis NATiV3 forecast model (40,000 draws); NATIVE, ESSENCE and MAESTRO-NASH publications (see Sources).

Exhibit 4

Forecast medians land at +10 to +15, below the +17 bar

NATiV3 1,200 mg placebo-adjusted composite gain against the three bars it will be judged on, with the two observed references, the bars the company has set and ten sensitivity runs of the forecast, in percentage points.

Statistical bar+3.8 to +6.8by placebo rateClinical floor+10ESSENCE-equivalent+17NATiV3 forecastWk 72 · modeled · 23% vs 11%Median +1280% range: 0 to +27NATIVE 1,200 mgWk 24 · observed · 31.3% vs 7.4%+23.9Positive,not differentiatedESSENCE (semaglutide)Wk 72 · NRI basis · 30.7% vs 13.6%+17Bars the company has setPowering back-solve · 9 Nov “win”Powering ~9 to 11Management +18Forecast sensitivity runsTen settings in A3 · modeledMedians +10 to +1580% ranges: −2 to +310+10+20+30Percentage points, 1,200 mg minus placebo
View data
ItemGain over placebo (pts)Basis
Statistical bar (p<0.05)+3.8 (5% placebo) to +6.8 (25%)Modeled; NRI, ~336/arm, two-sided 0.05
Clinical floor+10Placebo + 10 pts
ESSENCE-equivalent+17Placebo + 17 pts, NRI basis (cross-trial)
NATiV3 forecast, median+12Modeled; placebo 11%, 1,200 mg 23%
NATiV3 forecast, 80% range0 to +27Modeled; 40,000 simulated trials
NATIVE 1,200 mg, wk 24+23.9Observed; 31.3% vs 7.4%
Company powering assumptionabout +9 to +11Clinaptis back-solve of the 90% power claim (A1)
Management “win”+18CEO, 9 Nov 2026 conference
ESSENCE, wk 72+17Observed, NRI basis; 30.7% vs 13.6%
Sensitivity: v4 forecast assumptions, median / 80% range+12 (0 to +25)Modeled sensitivity run
Sensitivity: Equal weight, 5 programmes, median / 80% range+11 (−2 to +25)Modeled sensitivity run
Sensitivity: Positive programmes only, median / 80% range+15 (+6 to +29)Modeled sensitivity run
Sensitivity: No fibrosis allowance, median / 80% range+11 (0 to +23)Modeled sensitivity run
Sensitivity: Missing biopsies 15%, median / 80% range+13 (0 to +29)Modeled sensitivity run
Sensitivity: Missing biopsies 28%, median / 80% range+11 (0 to +25)Modeled sensitivity run
Sensitivity: Effect erodes (×0.85), median / 80% range+10 (0 to +22)Modeled sensitivity run
Sensitivity: Placebo near ESSENCE NRI, median / 80% range+13 (0 to +27)Modeled sensitivity run
Sensitivity: Placebo at inherited Base, median / 80% range+11 (0 to +23)Modeled sensitivity run
Sensitivity: NATIVE F2–F3 subset, median / 80% range+14 (0 to +31)Modeled sensitivity run

Statistical bar assumes non-responder imputation, ~336 per arm and two-sided 0.05; it rises from +3.8 at 5% placebo to +6.8 at 25% and is hypothetical because the SAP is not public. Forecast is the median of 40,000 simulated trials (placebo 11%, 1,200 mg 23%), with the 80% interval of the true gain. Sensitivity runs are the settings in A3.

Source: Clinaptis forecast models (modeled, not observed data); NATIVE and ESSENCE publications (see Sources).

03Benchmarks

How much histology would make lanifibranor competitive?

ESSENCE sets the closest benchmark at about 17 points on the composite and 15 on fibrosis. We forecast 12 and 13.

The central NATiV3 result clears the modeled statistical bar and falls short of the ESSENCE bar on both measures. The like-for-like ESSENCE figure is the share of patients with MASH resolution and fibrosis improvement together: 32.7% on semaglutide against 16.1% on placebo as published, a 16.5-point gain. Setting NATiV3’s composite against ESSENCE’s 63% resolution and 37% fibrosis rates compares separate active-arm response rates with a same-patient composite. Counting missing biopsies as failures, as we assume NATiV3 will, our conversion gives 30.7% against 13.6%, about 17 points. It reproduces ESSENCE’s published non-responder single-endpoint figures to within 0.2 points. ESSENCE is the closest cross-trial reference. Both trials read at 72 weeks, and F3 patients (69% against 67%) and type 2 diabetes (56% against 55%) make up similar shares, with NATiV3 on the company’s preliminary count.

On readout day investors will set a NATiV3 result near +12 against the +11.6 points MAESTRO-NASH reported for resmetirom. That comparison does not establish parity with Rezdiffra. The MAESTRO composite also required a two-point drop in disease activity score, was read at 52 weeks and included stage F1B patients. A result near +12 would tell us more about how much of NATIVE survived than about how lanifibranor compares with Rezdiffra.

Management has set its own bar higher. On 9 November the chief executive called 18 points on the primary “an absolute win” for F3 and diabetic patients, and on the September call the company framed 18 points of fibrosis as the result to repeat. Both amount to replicating Phase 2. The company’s own 90% power claim back-solves to a true composite gain of only about 9 to 11 points (A1), and our central +12 sits between the two. An observed gain of 18 or more comes up in about 30% of our simulations, so a result in our central band can be a clear statistical win and still read as short of management’s framing.

HARMONY shows how much stronger histology can look in a different study. Efruxifermin gained 28 points on the composite and roughly 30 on fibrosis in a 126-patient Phase 2b read at 96 weeks. We treat that as a higher observed result from a smaller trial with a different mechanism, and as no limit on what lanifibranor could show. Clearing our C-branch bar (roughly 30% in our model) would put lanifibranor at ESSENCE level on both measures. A statistically positive result can fall short of that bar, and even ESSENCE-level histology leaves the add-on case open (§4).

Cross-trial comparison

Exhibit 5

NATiV3 forecast falls short of ESSENCE on both histology measures

Gain over placebo in the composite (MASH resolution and fibrosis improvement in the same patient), in percentage points; the hollow marker is the fibrosis gain. All rows are non-cirrhotic MASH; NATIVE and MAESTRO-NASH also enrolled earlier-stage fibrosis. Rows run from most to least comparable to NATiV3, whose row is our forecast, not data.

ESSENCE: +17Trial, and comparability to NATiV3NATiV3 (lanifibranor 1,200 mg): Clinaptis forecast, composite +12 (80% range 0 to +27), fibrosis +13; readout pendingNATiV3 (lanifibranor 1,200 mg)Wk 72 · CRN · Clinaptis forecastForecast +12 (0 to 27)+13ESSENCE (semaglutide): composite +17 pts (30.7% vs 13.6%), fibrosis +15.4 pts, comparability highESSENCE (semaglutide)Wk 72 · CRN · 30.7% vs 13.6%High: same clock, 69% F3, 56% T2DComposite +17Fibrosis +15.4HARMONY (efruxifermin 50 mg): composite +28 pts (35% vs 7%), fibrosis +31 pts, comparability mediumHARMONY (efruxifermin 50 mg)Wk 96 · CRN · 35% vs 7%Medium: Phase 2b (n=126), weekly+28 observed+31MAESTRO-NASH (resmetirom 100 mg): composite +11.6 pts (16.0% vs 4.9%), fibrosis +11.8 pts, comparability lowMAESTRO-NASH (resmetirom 100 mg)Wk 52 · CRN + NAS ≥2 · 16.0% vs 4.9%Low: stricter definition, shorter+11.6+11.8NATIVE (lanifibranor 1,200 mg): composite +23.9 pts (31.3% vs 7.4%), fibrosis +18.7 pts, comparability lowNATIVE (lanifibranor 1,200 mg)Phase 2b anchor · Wk 24 · SAF-based · 31.3% vs 7.4%Low: 24% F0–F1, single reader+23.9+18.70102030Percentage points, active minus placebo
View data
TrialWeekDefinitionActive vs placeboComposite ΔFibrosis ΔComparability
NATiV3 (lanifibranor 1,200 mg)72CRN23% vs 11% (forecast)+12 (0 to +27)+13Forecast, not data
ESSENCE (semaglutide)72CRN30.7% vs 13.6%+17+15.4High: same clock, 69% F3, 56% T2D
HARMONY (efruxifermin 50 mg)96CRN35% vs 7%+28+31Medium: Phase 2b (n=126), weekly
MAESTRO-NASH (resmetirom 100 mg)52CRN + NAS ≥216.0% vs 4.9%+11.6+11.8Low: stricter definition, shorter
NATIVE (lanifibranor 1,200 mg)24SAF-based31.3% vs 7.4%+23.9+18.7Low: 24% F0–F1, single reader

NATiV3 row is our forecast median with its 80% range (Exhibit 3). ESSENCE composite is our non-responder conversion of the published missing-biopsy counts (published: 32.7 vs 16.1%, +16.5). HARMONY is the modified intention-to-treat population (n=126); fibrosis is 49 vs 19%, a published difference of 31 points. MAESTRO differences are as published. Reader variability alone moves placebo rates by about 4 points (Rezdiffra label). A needle marks an injectable drug and a pill an oral one.

Source: ESSENCE, MAESTRO-NASH, HARMONY and NATIVE publications; Rezdiffra label; Clinaptis forecast model and normalisation (see Sources).

04Add-on

What would make lanifibranor a useful add-on?

NATIVE’s tolerability was better than the class history suggests. The open question is whether a visible fluid signature stays acceptable over 72 weeks next to the fibrosis benefit.

NATIVE’s tolerability was better than the PPAR class history suggests. Discontinuations for adverse events ran at placebo levels (4% to 5% in every arm), with no muscle or kidney signal and improving liver enzymes, and HbA1c fell about half a point. The cost is a fluid signature. At 24 weeks, 1,200 mg added 2.7 kg and 800 mg added 2.4 kg against a 0.2 kg loss on placebo; a third of treated patients gained 5% or more of body weight, and NT-proBNP rose with dose and tracked the weight gain (cardiometabolic analysis). Peripheral oedema ran 8% and 6% against 2%, and anaemia 7% at 1,200 mg against 1% at 800 mg and none on placebo. The lower dose bought little relief on weight or oedema and gave up most of the fibrosis signal.

The company’s 48-week scleroderma data show weight plateauing around weeks 24 to 30, and management describes most oedema as mild to moderate. In early 2024 NATiV3 paused screening after a suspected unexpected serious liver reaction; the patient recovered, and enrolment resumed with more frequent liver monitoring and exclusion of patients prone to autoimmune liver or thyroid disease. Nine safety-board reviews have since recommended continuing. NATiV3 triples the exposure, so the 72-week weight trajectory, oedema, anaemia, heart-failure events and hepatic safety all matter.

Some payers (Neighborhood Health Plan of Rhode Island, a Medicaid plan, among them) already require a GLP-1 trial before Rezdiffra; others do not. Where lanifibranor follows a GLP-1, its job is residual disease. ESSENCE is why we look first at fibrosis: on a non-responder basis semaglutide added about 30 points on resolution but 15 on fibrosis. A strong 1,200 mg fibrosis result would give the add-on case something to build on. The GLP-1 subgroup can say whether that case looks plausible in the patients most likely to receive the drug.

About 14% of NATiV3 patients were on a stable GLP-1 at entry, roughly 47 an arm on our assumed arm size. With about 40 evaluable biopsies an arm, the subgroup gain carries a 95% interval of about ±16 points. It can flag a loss of effect but not prove one, and it cannot show the effect is retained, and patients who started a GLP-1 after randomisation are not a randomised comparison. The persuasive-but-undifferentiated value in §6 has to absorb that uncertainty, because a positive readout will not remove it.

IQVIA links slower US MASH trial enrollment (a median of 0.41 falling to 0.07 patients per site per month between 2012–15 and 2020–23) to rising GLP-1 use, though on few trials and across the pandemic years. The more direct reason only 14% of NATiV3 patients were on a GLP-1 is the trial’s stable-dose entry rule and enrolment that ran from 2021 to 2025, before Wegovy’s August 2025 approval in MASH. Either way, the trial may under-represent the GLP-1-treated patients who would receive lanifibranor in 2028.

05Financing

A positive headline can still put the debt in default

The warrants can open on a secondary endpoint. The debt requires the primary.

The T3 warrant terms make 77.3M shares at €1.50 (about €116M) exercisable once Inventiva releases topline data announcing that any key primary or key secondary endpoint has been met at any dose, by 15 June 2027. “Met” has no statistical definition, so the company’s wording decides. Holders then have 45 days and full discretion. The debt terms make failure of the primary composite an event of default. Curing it requires the key secondary endpoints to be met and the T3 warrants to be exercised in cash during a cure period whose length is not public.

The warrants open on the announcement; the cash arrives only if holders choose to pay

Financing terms that decide what a secondary-only result does to the equity.

T3 warrants

€116M

77.3M shares at €1.50

Trap-branch value per share

€1.40

Value without exercise, just below the strike

Exercise window after the release

45 days

Holders have full discretion

Funding deadline

15 Feb 2027

Tranche C (€55M) needs the primary and €100M or more from T3 or equity

Source: Inventiva 6-Ks of 15 Oct 2024 and 3 Jun 2026 (financing terms); Clinaptis valuation model (Trap-branch value, modeled). See Sources.

Take the case where the primary misses and a secondary is met. The release opens the warrants and defaults the debt on the same day. In our Trap branch the shares are worth €1.40, just below the €1.50 strike, so exercise is not a simple market decision. The largest T3 holders, NEA, BVF and Samsara among them, also hold large positions in shares and pre-funded warrants, and paying a little above value to cure the default can protect those stakes. The cure also requires the key secondary endpoints, plural, to be met, which a secondary-only result may not satisfy. We treat €1.40 as the value without exercise, within a judgment range of about €1.00 to €2.00, skewed towards €1.00, depending on whether the cure succeeds and how the market reads a near miss; the weighted return uses €1.40. Without the cash the cure fails, lenders can accelerate, and the equity falls further. Exercisability is not cash: the warrants open on the announcement, but the money arrives only if holders choose to pay.

A primary hit avoids the default but not the funding deadline. Tranche C (€55M) and the extension of Tranche B’s interest-only period both need the primary composite and at least €100M from T3 exercise or new equity by 15 February 2027. A late-December topline would leave the 45-day exercise window running up to that date.

Note. Holders of 60% of the T3 warrants can waive the trigger on a change of control. During the cure period the company must hold cash equal to 100% of the A and B principal in its Luxembourg accounts. After results, holders can demand partial prepayment if the A and B principal exceeds 10% of market capitalisation. If the default is cured through T3, a mandatory prepayment to 7.5% of post-cure market capitalisation applies, and below €400M (calculated under non-public issue documents) holders can demand full repayment. Cash and deposits were €233.9M at 30 June, with runway guided to the end of Q2 2027 and going-concern language retained. Q3 cash is due on 23 November.

06Valuation

What an ordinary win has to be worth

At €3.36, a clinically persuasive but undifferentiated result (branch B) needs to support about €0.95B of enterprise value for the probability-weighted trade to break even. Inventiva trades at about €0.96B of enterprise value today. Our B value is €1.6B, and we test €1.0B to €2.0B. Across that range the expected return, after a modeled deduction for the royalty certificates, runs from about +1% to +16% (+10% at €1.6B). About one outcome in five loses more than half the share price, and about half gain a quarter or more. Branch values sit between −82% and +88% of the price, so this is the spread of the weighted value and no forecast of the readout-day move. Neutral, with the skew to the upside.

Valuation

Exhibit 6

Value per share by NATiV3 outcome: above the price, with deep tails

Bar width is the model probability of each outcome; bar height is our post-data value per share, with the persuasive-but-undifferentiated branch (B) at €1.6B of enterprise value. The weighted value is €3.70 against €3.36 at the 10 November close, and equals the price when B is worth €0.95B.

€0 €2 €4 €6 €8 €6.31C · 30%Best case: ESSENCE-level€4.24B · 23%Persuasive€2.62A · 25%Narrow hit€1.40Trap · 10%Secondaries€0.61Miss · 13%Nothing met Price €3.36 Weighted €3.70 T3 strike €1.50 Share of probability (bar width), Clinaptis model
View data
BranchModel weightFinancing consequenceAssumed EVValue / share
C: ESSENCE-level30%T3 likely exercised (+€116M); Tranche C available; CTTQ $10M€2.5B (−€0.25B royalty)€6.31 (+88%)
B: persuasive, not differentiated23%As above€1.6B (−€0.16B)€4.24 (+26%)
A: narrow hit (composite under 10 or weak fibrosis)25%As above; partnering leverage falls€0.9B (−€0.09B)€2.62 (−22%)
Trap: secondaries only10%Event of default; exercise depends on holders’ other stakes (§5); repayment tests€0.35B€1.40 (−58%)
Miss: nothing met13%Default; debt repaid from cash€0.1B€0.61 (−82%)
Break-even value of branch B at the 10 November price: €0.95B. Weighted return at a B value of €1.0B / 1.2B / 1.3B / 1.6B / 2.0B: +1% / +4% / +5% / +10% / +16%.

Branch weights are model outputs, rounded independently; across the A3 settings C runs from 20% to 37% and Miss from 2% to 20%. Per share on 313.7M economic-basic shares, or 391.0M after T3 (A5). Net cash at readout is an estimate of €90M. EVs are judgments about post-data trading, not fundamental values.

Source: Inventiva H1 2026 report, 6-Ks and FY2025 20-F; Euronext close 10 Nov 2026; Clinaptis model and estimates (modeled, not observed data).

Our edge on the trial is about ten points of probability, and it depends on B. We put a primary hit at 78%. On our branch values the price breaks even at 68%, but that break-even moves from 77% to 63% as B moves from €1.0B to €2.0B, so the edge is as much a statement about B as about the trial. It is our values restated against the price, not a reading of market expectations. Across our forecast settings the return at €1.6B runs from −2% (erosion) to +27% (positive programmes only; Appendix A3). B is still the assumption to argue with, and so is C: a 30% weight on €6.31 now carries about half the weighted value.

In sales terms, B at €1.6B needs roughly $2B of peak sales and the €0.95B break-even about $1.3B; the €2.0B top of our range needs about $2.6B (Appendix A4, extrapolated). Our model peaks in 2034. Rezdiffra, the first approved MASH drug, is running at about $1.7B annualised off Madrigal’s Q3 sales of $431M. The break-even therefore asks an oral add-on behind GLP-1s to peak at about three-quarters of where the category leader runs today, and our B value asks it to exceed that. That is a deliberate tension: B prices a Rezdiffra-scale franchise for a result our forecast puts below ESSENCE, and it is the main way this note could be wrong. Patient numbers are unlikely to bind: modelling by Estes and colleagues, cited by IQVIA, projects about 14M US patients with F2–F4 disease by 2030, and Inventiva estimates 374K F2–F3 patients under treater care in 2025 (the low end of published estimates), against more than 55,850 on Rezdiffra. At Rezdiffra’s roughly $31,000 of net revenue per patient on therapy, B’s peak needs about 70,000 paid patient-years. We infer that diagnosis, payer steps and persistence set the pace.

Precedent brackets the value without settling it. Madrigal added about $2.9B in a day on MAESTRO-NASH as the first approved drug. Novo paid $4.7B in cash plus a $0.5B CVR for Akero, and Roche $2.4B upfront for 89bio, both for fibrosis-led FGF21 injectables. Genfit lost about half its value when elafibranor missed. None is a standalone oral PPAR with weight gain, so our range of €1.0B to €2.0B is a judgment. The royalty certificates take a further share of sales. We model them as debt, about €0.25B in the best branch and scaled down in A and B, which costs roughly ten points of expected return.

Our branch values assume a US launch in mid-2028. FDA’s January 2025 draft guidance, under which FDA generally intends to require a confirmatory trial to be underway before accelerated approval, is a separate risk to that date, since Inventiva’s confirmatory trial, in compensated cirrhosis, is planned to start in 2027. The trade works if a persuasive but ordinary result can carry an oral add-on to roughly Rezdiffra’s current scale, after the certificates and on a 2028 launch.

07Madrigal

Madrigal: a clean IVA win adds a rival; a miss changes little

A clean NATiV3 win adds an oral rival to our Negative on Madrigal. A miss leaves the existing GLP-1 and FGF21 pressures intact.

Our June note on Madrigal argued that Rezdiffra is being squeezed between GLP-1 step edits on one side and FGF21 drugs on the other. It did not discuss lanifibranor. Fundamentals have since run ahead of that bear case (Q3 sales of $431M, more than 55,850 patients, a first profit), so the case now rests on competitive sentiment. Our model assumes a 2028 lanifibranor launch, which makes NATiV3 a sentiment event for MDGL and no earnings event. The reactions in Exhibit 7 are judgments, not event-study estimates.

Exhibit 7

NATiV3 outcome and the read-through to our Madrigal Negative

NATiV3 outcomeExpected MDGL reactionWhy, and what it does to our Negative
Miss or high-placebo failUp, small, likely fadedRemoves a 2028 oral rival but leaves the GLP-1 and FGF21 squeezes intact. Neutral to slightly challenging.
Hit at +10–17, weight gainDown, smallA third F2–F3 candidate, but a niche PPAR behind GLP-1 and Rezdiffra in payer steps. Mildly supportive.
Hit at ≥+17, fibrosis ≥+15, GLP-1 subgroup not materially below overallDown, meaningfulAn oral insulin sensitiser competes inside the GLP-1-treated residual population, and is skewed to T2D, 68% of MAESTRO’s population. Supportive.
Bull, then Madrigal acquires IVAMixedNeutralises a rival and adds an all-oral combination partner. A path that challenges our Negative.

Magnitudes are judgments, not event-study estimates. Expect Madrigal to answer any cross-trial claim with re-analyses on aligned endpoints, as it did after ESSENCE.

Source: Clinaptis MDGL note (3 Jun 2026); Madrigal Q3 2026 release (3 Nov 2026); Clinaptis estimates.

08Readout

What to read in the press release, in order

  1. Wording. A release that says a key secondary endpoint was met and is silent on the primary is a reason to check, and no confirmed miss. We call the trap branch only once the primary is confirmed missed and a secondary confirmed met. Then separate the two financing effects: the warrants become exercisable on the announcement, and cash arrives only if holders pay €1.50.
  2. Composite gain at 1,200 mg. Subtract placebo from the 1,200 mg response rate first: every bar here is a gain. Below about 7 points is the lower tail of our range, close to the statistical line. 10 to 17 is our central band. 17 or more matches ESSENCE, and 18 is management’s own win line.
  3. Fibrosis gain at 1,200 mg, set against the composite. Read it as a ratio. In NATIVE, 1,200 mg delivered 0.78 points of fibrosis gain per point of composite and 800 mg only 0.33; ESSENCE ran about 0.9 and MAESTRO-NASH about 1.0. Below about +6 the result is resolution-led, +6 to +12 is Rezdiffra-like, and +15 or more is ESSENCE-like. Our central fibrosis gain is +13.
  4. Then sort the branch, top down (rules as in Appendix A3). C needs the primary met at either dose, a 1,200 mg composite gain of 17 or more, a 1,200 mg fibrosis gain of 15 or more and both 1,200 mg key secondaries significant. B needs the primary met, a 1,200 mg composite gain of at least 10, a fibrosis gain of at least 10 and both 1,200 mg key secondaries (resolution and fibrosis) significant, but not C. Any other met primary is A. Fibrosis, significant at 1,200 mg in about 75% of our simulations, is the usual swing between B and A. If the release leaves out a figure a branch needs, treat that branch as provisional: a +17 composite on its own establishes only the composite benchmark. Within fibrosis, the F3 subgroup (two-thirds of patients) and two-stage regression are where FGF21 data set the comparison.
  5. Placebo composite rate and biopsy completion. Our 80% placebo range is 6% to 18%. The statistical bar rises with it, from about 4 points at 5% to about 7 at 25%, but fixing placebo near ESSENCE left our hit rate unchanged (§2). We assume 22% missing biopsies; each 5 points more takes about a point off the composite gain. Check the analysis population and both active arms before reading anything into a high rate.
  6. 800 mg. A significant composite would be better than our central forecast (+4 points, nominally significant in about 40% of simulations). A fibrosis signal at 800 mg would be a genuine surprise.
  7. Weight, oedema, haemoglobin and heart-failure events at 72 weeks, placebo-adjusted. A weight gain above about 4 kg is where we would reassess tolerability and adoption. It is a threshold for our view, and no prediction of the label.
  8. Liver stiffness and blood markers. In NATIVE, liver stiffness fell only about 1 kPa against placebo, which management attributes to fluid retention. Real-world prescribing and payer checks run on such non-invasive tests, so a weak 72-week stiffness signal would matter for adoption even after a histology win.

What would change our mind

  • Publication of the NATiV3 SAP showing a missing-data rule, GLP-1 estimand or multiplicity scheme materially different from our assumptions.
  • Disclosure of how the biopsies were read (single reader, multi-reader consensus or AI-assisted), which would change how we read placebo response and the width of the forecast range. The direction depends on what the protocol specifies.
  • Any 48–72-week weight or oedema data, or combination histology with a GLP-1.
  • Regulatory: FDA’s January 2025 draft guidance, under which FDA generally intends to require confirmatory trials to be “underway” before accelerated approval. NATiV3’s 48-week extension is a safety study, and the outcomes trial in compensated cirrhosis is not planned to start until 2027, so 2028 launch timing is at risk independent of the data.
  • Disclosure of the final biopsy completion rate, against our 22% missing-biopsy assumption.
AAppendix

How the forecast is built and how sensitive it is

A1Forecast method: placebo rates, Phase 2-to-3 translation, missing biopsies and trial simulation

The forecast has four layers. First, week-72 placebo rates for MASH resolution and fibrosis improvement are drawn from a prior anchored on 72-week F2–F3 placebo arms on a non-responder basis (ESSENCE 28.9% and 18.8%; semaglutide Phase 2 17% and 33%), with a spread wide enough to cover reader and population differences. Second, each dose’s week-72 effect on the odds scale equals NATIVE’s observed effect (with its sampling error) multiplied by a Phase 2-to-3 translation factor drawn from the five-programme calibration (A2), failures at half weight; for fibrosis the factor is scaled up by a quarter to allow for the longer treatment period (§2). The composite is built from the two endpoints and their observed association (NATIVE and ESSENCE, similar in both arms), not by multiplying them. Third, every arm’s response rates are scaled for 22% missing biopsies, against about 15% in the anchors. Fourth, one NATiV3 is simulated per draw (about 336 per arm, shared placebo, Hochberg across doses, non-responder imputation) and classified into the five branches. Arm sizes and the testing scheme are assumptions; the SAP is not public. The company describes 90% power “assuming higher placebo response and lower efficacy” than Phase 2b; at about 336 an arm that back-solves to a true composite gain of roughly 9 to 11 points (company-framed and our back-solve, a rough cross-check because the assumed placebo rate is unknown).

A2Phase 2-to-Phase 3 efficacy translation in MASH trials

Three of five programmes kept at least three-quarters of their Phase 2 effect. Elafibranor kept 0.59 and failed, and cenicriviroc reversed. The house view gives those two failures half weight (A3).

ProgrammeEndpointPhase 2 act / pboPhase 3 act / pboTranslationOutcome
InjectableSemaglutideResolution; composite59 / 17
37 / 15
58.8 / 28.9
30.7 / 13.6
0.75
Approved (ESSENCE)
OralResmetiromResolution24.7 / 6.525.9–29.9 / 9.7
0.82
Approved (MAESTRO-NASH)
OralObeticholic acidFibrosis35 / 1923 / 12
0.94
Positive Phase 3; not approved
OralElafibranorResolution19 / 1219.2 / 14.7
0.59
Failed (RESOLVE-IT)
OralCenicrivirocFibrosis20 / 1022.3 / 25.5
−0.22
Failed (AURORA)

Translation is the Phase 3 log-odds ratio divided by the Phase 2 ratio, averaged over matched endpoints with a nominally significant Phase 2 signal. Bars run from −0.4 to 1.0, where 1.0 is no loss of effect; red marks the two failed programmes. Cenicriviroc’s negative value reflects a Phase 3 placebo arm that outperformed active, not only lost effect. A needle marks an injectable drug and a pill an oral one.

A3Branch probabilities and expected return under eleven settings

The house view adds a fibrosis allowance for the longer trial and a 22% missing-biopsy assumption to the v4 forecast (§2). Each 5 points of missing biopsies above that takes about a point off the composite gain; final biopsy completion is not disclosed.

SettingPrimary metCBATrapMissWeighted return
House view: failures at half weight, fibrosis allowance, 22% missing78%30%23%25%10%13%+10%
v4 forecast assumptions (no allowance, 15% missing)76%25%22%29%11%13%+4%
All five programmes equal weight70%25%21%25%10%20%+0%
Positive programmes only90%37%26%27%8%2%+27%
No fibrosis allowance75%21%22%31%12%14%−1%
Missing biopsies 15% (ESSENCE-like)79%34%22%23%9%12%+15%
Missing biopsies 28%76%25%23%28%10%14%+5%
Effect erodes (translation ×0.85)73%20%23%30%13%14%−2%
Placebo fixed near ESSENCE78%31%22%24%10%13%+12%
Placebo at inherited Base (15% / 20%)77%22%25%30%10%13%+3%
Re-anchored on NATIVE F2–F3 subset80%32%18%30%9%11%+12%

Branch rules: C = primary met, 1,200 mg composite gain ≥17, fibrosis gain ≥15 and both 1,200 mg key secondaries p<0.05; B = primary met, composite gain ≥10, fibrosis gain ≥10 and both key secondaries p<0.05, not C; A = primary met otherwise; Trap = primary missed but a key secondary met at any dose; Miss = nothing met. Shading is a column colour scale; weighted return is green above zero and red below.

Weighted return uses the Exhibit 6 branch values (B at €1.6B) and the €3.36 close. At a B value of €1.0B the settings give −12% (erosion) to +16% (positive programmes only), and at €2.0B +4% to +34%. The headline +1% to +16% holds the house-view odds fixed and moves B from €1.0B to €2.0B.

A4Peak sales needed to justify each branch value

At a 45% peak margin our €1.6B value for the persuasive branch needs about $2.1B of peak sales, 1.2 times Rezdiffra’s current run-rate of about $1.7B. The €2.0B top of our range needs about $2.6B (extrapolated).

Peak operating marginC: €2.5BB: €1.6BBreak-even B: €0.95BA: €0.9B
40%$3.6B$2.3B$1.4B$1.3B
45% (base)$3.3B$2.1B$1.3B$1.2B
50%$3.0B$1.9B$1.2B$1.1B

Peak sales that make a simple franchise DCF equal each branch value: US launch mid-2028, peak in 2034, flat to 2039 and then eroding, 25% tax, 85% probability of approval after a met primary, discounted at the Clinaptis 10.5% to end-2026 and converted at $1.17 per euro. Values are gross of the royalty certificates (A5); the break-even column is interpolated between the A and B values. Rezdiffra’s run-rate is four times Madrigal’s Q3 2026 net sales of $431M.

A5Price bridge: shares, cash, royalty certificates and break-even odds

Share bridgeShares (M)Exercise price
Ordinary shares, 31 Aug 2026236.28—
Pre-funded warrants (2024–25 financing)61.75Prepaid
New EIB warrants (July 2026)15.68€0.01
Economic basic313.70
T3 warrants77.33€1.50
After T3391.03

Values per share use the economic-basic count, or the count after T3 where T3 is exercised. The pre-funded and EIB warrants together (77.4M) happen to almost equal the T3 count (77.3M); they are different instruments. The 30 September ordinary count of 237.99M would lower values by about 0.5%; lender warrants and 29.8M employee instruments are excluded. We assume net cash of €90M at readout, an estimate pending the Q3 report on 23 November, and €215M after T3 (€116M) and the CTTQ milestone ($10M).

The royalty certificates carry a 2% royalty on US, EU and UK net sales capped at €92.1M, plus a separate 3% royalty with no cap disclosed in the 20-F text we read; the undiscounted cash requirement was €536.8M at end-2025. We treat them as debt: the full amount paid evenly over 2029–2042 and discounted at 10.5% is €0.25B, deducted in C and scaled by branch value in B and A.

The break-even probability holds the model mix within the success branches (C, B, A) and within the failure branches fixed, then solves for the odds of a primary hit that set the weighted value equal to the €3.36 price. That gives about 68% at a B value of €1.6B, and 77% to 63% as B moves from €1.0B to €2.0B. It restates our branch values against the price; it is not a market-implied probability. Scripts: iva_step7_valuation_v2.py, iva_step8_implied.py and iva_step9_forecast_v5.py.

A6Share price against US biotech

Exhibit A6

Inventiva’s ADS is up 6% in three years while US biotech has more than doubled

Daily closes of the Nasdaq-listed ADS and the XBI biotech ETF, indexed to 100 on 13 Nov 2023. The ADS fell to $1.65 on 9 Oct 2024, peaked at $7.15 on 23 Jan 2026 and closed at $3.95 on 10 Nov 2026; the Euronext share closed at €3.36 that day.

50100150200250Jan 2024Jul 2024Jan 2025Jul 2025Jan 2026Jul 202614 Oct 2024Structured financing15 Aug 2025Wegovy MASH approval2 Jun 2026RecapitalisationXBI 230 +130%IVA ADS 106 +6%124-ptgap

Swipe sideways to see the full chart.

The ADS represents one Euronext ordinary share. Closes are in US dollars, so the line includes euro/dollar moves, and US and Paris close at different times. Markers show announcement dates; the Wegovy approval was announced after the US close. Price performance excludes distributions.

Source: S&P Global Market Intelligence via Stock Analysis (daily closes); event dates from Inventiva, Novo Nordisk and FDA releases (see Sources).

SReferences

Sources

Trials, labels and registry

  • Francque SM et al. Lanifibranor in NASH (NATIVE). NEJM 2021;385:1547, with Supplementary Appendix, protocol and statistical analysis plan. NEJM.
  • Sanyal AJ et al. Semaglutide in MASH (ESSENCE). NEJM 2025, with Supplement Table S4. NEJM.
  • Harrison SA et al. Resmetirom in MASH (MAESTRO-NASH). NEJM 2024, Table 2. NEJM.
  • Lanifibranor cardiometabolic analysis of NATIVE, Nature Communications 2024 (PMC). Estes C et al., Hepatology 2018;67:123 (disease-burden model).
  • Efruxifermin in MASH (HARMONY), 96-week results. Lancet 2025;406:719. PubMed.
  • Programmes in A2: Newsome et al., NEJM 2021; Harrison et al., Lancet 2019; Neuschwander-Tetri et al., Lancet 2015; Younossi et al., Lancet 2019; Ratziu et al., Gastroenterology 2016; Genfit Form 6-K, 11 May 2020; Friedman et al., Hepatology 2018; Anstee et al., CGH 2024.
  • Rezdiffra US prescribing information (FDA, July 2026). FDA label. NATiV3 record: ClinicalTrials.gov NCT04849728. FDA draft guidance on confirmatory trials underway (January 2025); EMA reflection paper on NASH (2024).

Company filings, markets and payers

Clinaptis models

  • NATiV3 forecast (40,000 draws, 6 Nov 2026; V5 house view of 12 Nov 2026 adds the fibrosis allowance and missing-biopsy assumption), valuation and break-even models; scripts and outputs are listed under Companion files. Modeled, not observed data.