$1,362,000,000
Success-case NPV
44%
IRR
7.0%
PoS at Gate 0
$43M
NPV swing from a 20% cost overrun
Contents
1. The Case, and the Number That Undermines It
Presented to the Development Committee at Gate 4, 30 June 2026, in support of releasing $125,800,000 — 58% of the program, and the largest single commitment it will make.
| Value | |
|---|---|
| Success-case NPV, conditional on approval | $1,362,000,000 |
| Internal rate of return | 43.6% |
| Peak annual revenue | $1,640,000,000 |
| Probability of technical and regulatory success at Gate 0 | 7.0% — 47% × 30% × 50% |
| ⚠ Risk-adjusted expected value at Gate 0 | $96,000,000 |
| Authorized program cost | $243,040,000 |
At Gate 0 the risk-adjusted expected value of this program was
$96,000,000 against an authorized cost of $243,040,000. The expected value was less
than the cost of finding out.
Read as a conventional investment appraisal, that says do not proceed. Read correctly, it says something different: at candidate selection the company was not buying a product. It was buying the option to keep going. A 7.0% probability applied to a large number produces a small number, and no amount of confidence changes that arithmetic.
The response is not to inflate the forecast. It is to release the money one stage at a time, so that the largest commitment is made against the best evidence — which is exactly what this gate is doing.
Read as a conventional investment appraisal, that says do not proceed. Read correctly, it says something different: at candidate selection the company was not buying a product. It was buying the option to keep going. A 7.0% probability applied to a large number produces a small number, and no amount of confidence changes that arithmetic.
The response is not to inflate the forecast. It is to release the money one stage at a time, so that the largest commitment is made against the best evidence — which is exactly what this gate is doing.
2. Why the Number Moves Without the Asset Changing
| Gate | Risk-adjusted NPV | Movement |
|---|---|---|
| Gate 0 — candidate selection | $96,025,247 | |
| Gate 3 — Phase 2 readiness | $204,309,036 | ×2.1 vs Gate 0 |
| Gate 4 — Phase 3 initiation (today) | $681,030,120 | ×7.1 vs Gate 0 |
| Gate 5 — submission | $1,253,095,420 | ×13.0 vs Gate 0 |
The molecule is no better today than it was at Gate 0. The value has multiplied because two of
the three ways it could have failed are now behind us.
Risk adjustment is a convention applied at a point in time. At Gate 0 the figure carried the full 7.0% probability chain. At Gate 4 the Phase 1 and Phase 2 transitions have resolved, so only the Phase 3 and approval risk remains — and the same forecast, differently adjusted, is worth several times more.
Which means comparing a Gate 0 figure with a Gate 4 figure and calling the difference “value created” is a category error. Nothing was created. Uncertainty was removed, which is what the money bought.
Risk adjustment is a convention applied at a point in time. At Gate 0 the figure carried the full 7.0% probability chain. At Gate 4 the Phase 1 and Phase 2 transitions have resolved, so only the Phase 3 and approval risk remains — and the same forecast, differently adjusted, is worth several times more.
Which means comparing a Gate 0 figure with a Gate 4 figure and calling the difference “value created” is a category error. Nothing was created. Uncertainty was removed, which is what the money bought.
The corresponding discipline is that the success case must never be quoted without its condition. $1,362,000,000 is the value if approved. Quoted at Gate 0 without the qualifier it overstates the expectation roughly fourteenfold, and quoting it that way is the most common way a development program is oversold internally.
3. The Assumption Stack
Peak revenue is a chain of five multiplications. Every link is an assumption, and errors in them compound rather than average.
| Assumption | Value | What it rests on |
|---|---|---|
| Eligible US population | 41,000,000 | Adults meeting the BMI criteria. Epidemiological, stable, and the least contested number in the chain. |
| Treated fraction | 4.0% | ⚠ The proportion who will be on ANY drug in this class at peak. A market-growth assumption, not a share assumption, and the one with the widest credible range. |
| Our share of treated patients | 8% | Fourth entrant into a class where two incumbents hold preferred formulary position. Assumes a differentiating claim we do not yet have. |
| Net price per patient-year | $12,500 | ⚠ WAC less an assumed 26% gross-to-net erosion. The assumption the CFO abstained over, un-refreshed since the original case. |
| Persistence at 12 months | 80% | Chronic therapy. Every revenue model in this class is a persistence model wearing a revenue model's clothes. |
Two of these five are load-bearing and neither is settled.
The share assumption presumes a differentiating claim the program does not yet have — it rests on a tolerability endpoint sitting fourth in a six-position testing hierarchy. And the net price assumes 26% gross-to-net erosion on a figure that has not been refreshed since the original case.
Multiply a 30% share miss by a 10% price miss and peak revenue falls by more than a third, which is what makes an assumption stack different from a list of assumptions. They do not offset.
The share assumption presumes a differentiating claim the program does not yet have — it rests on a tolerability endpoint sitting fourth in a six-position testing hierarchy. And the net price assumes 26% gross-to-net erosion on a figure that has not been refreshed since the original case.
Multiply a 30% share miss by a 10% price miss and peak revenue falls by more than a third, which is what makes an assumption stack different from a list of assumptions. They do not offset.
The most contested figure is the treated fraction. 4.0% of the eligible population being on any drug in this class at peak is a market-growth assumption rather than a competitive one — and it is the number over which reasonable people disagree by a factor of two, because it depends on payer behavior across an entire category rather than on anything this program does.
4. Sensitivity
One driver at a time, against the success-case NPV.
| Driver | Low | Base | High | NPV swing |
|---|---|---|---|---|
| Share of treated patients | 5% | 8% | 11% | ██████████████ $512M |
| Net price per patient-year | $10,500 | $12,500 | $13,500 | ███████████ $386M |
| Treated fraction | 3.0% | 4.0% | 5.0% | █████████ $341M |
| Persistence at 12 months | 70% | 80% | 85% | ██████ $214M |
| Launch timing | +12 months | on plan | — | ███ $118M |
| Program cost | +20% | on plan | −10% | █ $43M |
Program cost is the least sensitive driver in the entire model, by roughly twelvefold.
A 20% cost overrun — $43,400,000, an outcome that would dominate every status report for a year — moves NPV by about $43M. A three-point share miss moves it by $512M.
A program managed primarily to budget is optimizing the smallest lever it has. That is not an argument for spending carelessly; the ceiling is real and the Board authorized it. It is an argument about attention — and about the fact that the things which decide this program's outcome are largely decided outside the program's cost baseline.
A 20% cost overrun — $43,400,000, an outcome that would dominate every status report for a year — moves NPV by about $43M. A three-point share miss moves it by $512M.
A program managed primarily to budget is optimizing the smallest lever it has. That is not an argument for spending carelessly; the ceiling is real and the Board authorized it. It is an argument about attention — and about the fact that the things which decide this program's outcome are largely decided outside the program's cost baseline.
This assert is in the fact base, because it is the argument the whole suite rests on: if program cost were ever the largest driver, this would be a cost-control program rather than an evidence-generation program, and almost every methodological choice in the other fifty-seven artifacts would be wrong.
5. What Changed Since Gate 3
| Element | Direction | What happened |
|---|---|---|
| Probability of success | de-risked | Phase 2 survived, so the Phase 2→3 transition probability no longer applies. Risk-adjusted value rises without the asset improving. |
| Scope | CR-02 added | A cardiovascular sub-study, $6,400,000, pre-empting a likely post-marketing requirement. |
| Cost | contingency drawn | The March analytical method transfer failure. Absorbed within contingency; ceiling unchanged. |
| Tolerability evidence | weaker than target | ⚠ Phase 2 GI discontinuation 5.1% against a ≤4% target and a ≤7% minimum. Clears the minimum, misses the target, and it is the differentiator. |
The tolerability result is the one that matters, and it is presented as it is rather than as
the program would like it.
Phase 2 gastrointestinal discontinuation came in at 5.1% against a ≤4% target and a ≤7% minimum. It clears the minimum, so the program proceeds. It misses the target, and the target is the differentiator.
A business case that reported “tolerability within acceptable limits” would be accurate and would have concealed the only genuinely bad news in the package. Clearing a minimum is not the same as meeting a target, and a case that blurs the two has made the gate decision harder rather than easier.
Phase 2 gastrointestinal discontinuation came in at 5.1% against a ≤4% target and a ≤7% minimum. It clears the minimum, so the program proceeds. It misses the target, and the target is the differentiator.
A business case that reported “tolerability within acceptable limits” would be accurate and would have concealed the only genuinely bad news in the package. Clearing a minimum is not the same as meeting a target, and a case that blurs the two has made the gate decision harder rather than easier.
6. Where the Case Is Soft
| Not claimed | Why not |
|---|---|
| That the program is worth its cost on expected value | It is not, and was not at Gate 0. Expected value then was 96M against an authorized 243M. ⚠ The case is for the OPTION, not the asset. |
| That the share assumption is defensible today | It rests on a differentiating claim that sits fourth in a six-position testing hierarchy and has not been tested. Gate condition GC-03 exists because of it. |
| That the gross-to-net assumption is current | ⚠ It is not. It has not been refreshed since the original case, which is the recorded basis of the Chief Financial Officer's abstention at this gate. |
| That cost control determines the outcome | Program cost is the LEAST sensitive driver in the model — a 20% overrun moves NPV less than a 3-point share miss. ⚠ A program managed primarily to budget is optimizing the smallest lever it has. |
A business case is more useful for what it refuses to assert than for what it asserts.
Everything in the left-hand column is something the program would prefer to be able to say, and the reason for writing them down is that a Committee reading only the favorable case has no way to know which parts are soft. Stating the softness explicitly is what makes the rest of the document usable as evidence rather than as advocacy.
The third row is the live one at this gate. The gross-to-net assumption has not been refreshed, the program knows it has not, and the Chief Financial Officer has recorded an abstention on precisely that basis rather than voting against a program he believes should proceed.
Everything in the left-hand column is something the program would prefer to be able to say, and the reason for writing them down is that a Committee reading only the favorable case has no way to know which parts are soft. Stating the softness explicitly is what makes the rest of the document usable as evidence rather than as advocacy.
The third row is the live one at this gate. The gross-to-net assumption has not been refreshed, the program knows it has not, and the Chief Financial Officer has recorded an abstention on precisely that basis rather than voting against a program he believes should proceed.
That abstention is the most efficient governance act in this program's history. It did not block anything, and it made a soft assumption impossible to lose track of — it now appears in the charter, the gate decision record, the risk report, the labeling strategy and here. A vote against would have been overruled and forgotten; an abstention with a stated reason survives.
7. The Recommendation
| Position | |
|---|---|
| Recommendation | Proceed — release $125,800,000 for Stage 4 |
| On what basis | Phase 2 delivered a dose and an efficacy signal sufficient to support two pivotals. The remaining risk is Phase 3 execution and approval, which is the risk this stage exists to retire. |
| What the Committee is NOT being asked | ⚠ To approve the commercial forecast, the label, or the launch. This gate authorizes a tranche and a Phase 3 program. Nothing further. |
| Conditions proposed | CMC readiness to a defined standard before Gate 5; the CV sub-study protocol issued to the DMC; a market access evidence plan approved and executed. |
The honest summary of this case: the science supports proceeding and the commercial position
does not yet support the forecast.
Those are different questions with different evidence and different timelines, and a gate that conflates them either stops a good program on a soft number or waves through a soft number because the science is good.
The conditions attached are the mechanism for holding both at once — proceed on the science, and require the commercial evidence to be built while the trials run, so that the question can be answered properly at Gate 5 rather than assumed here.
Those are different questions with different evidence and different timelines, and a gate that conflates them either stops a good program on a soft number or waves through a soft number because the science is good.
The conditions attached are the mechanism for holding both at once — proceed on the science, and require the commercial evidence to be built while the trials run, so that the question can be answered properly at Gate 5 rather than assumed here.
What this case could not do, and did not attempt, was create the years of payer evidence collection that would have made the access assumption solid. That work needed to begin in Stage 2. A gate can require evidence; it cannot retroactively create the time in which it should have been gathered.