$IBRX ImmunityBio: the bladder cancer with no approved treatment, and the regulatory window that narrowed

01What non-muscle-invasive bladder cancer is

Bladder cancer is common, survivable, and expensive in a way few other cancers are. The National
Cancer Institute’s SEER program estimates 84,530 new cases and 17,870 deaths in the United States in 2026. Five-year relative survival is 79.1%, far above most solid tumors. The
median age at diagnosis is 73, and three quarters of diagnoses happen after 65.

The reason it costs so much is that it keeps coming back. About three quarters of patients are
diagnosed with non-muscle-invasive disease — the tumor sits in the lining of the bladder and has
not yet grown into the muscle wall. That is good news for survival and bad news for everything else,
because the standard of care becomes a decade of surveillance: a camera into the bladder every three
months, then every six, biopsies whenever something looks wrong, and a new treatment every time the
disease returns. A systematic review published in European Urology in December 2024 found that
high-risk non-muscle-invasive disease that progresses after BCG runs past $200,000 in costs over
five years.

So the clinical problem is not usually how to keep the patient alive this year. It is how to keep
the disease out of the muscle wall, and how to keep the bladder in the patient, for as long as
possible.

02CIS and papillary: the distinction that decides everything

Two words decide which patients a drug is allowed to treat, and almost every regulatory event in
this field turns on the difference between them.

Papillary tumors grow outward, like small fronds or mushrooms attached to the bladder
lining by a stalk. In staging language they are Ta when they stay in the lining, and
T1 when they have grown into the connective tissue underneath. A urologist can see them
through a cystoscope and can usually cut them out.

Carcinoma in situ, universally abbreviated CIS and formally staged as Tis, is
not a lump at all. It is a flat sheet of malignant cells spreading across the surface of the bladder
lining. It is high grade by definition, it is hard to see, and it cannot be reliably resected,
because there is no mass to remove. Two patients can have the same stage of non-muscle-invasive
disease and completely different problems.

Here is why that matters commercially, and it is the single most important fact here.
Every drug approved for BCG-refractory bladder cancer in the modern era is approved only for
patients who have CIS — with or without papillary tumors alongside it. A patient whose disease
is papillary only, with no CIS, has no approved option anywhere in the world. That population is
what ImmunityBio has spent two years trying to reach.

03BCG: a tuberculosis vaccine that treats a tumour

The standard treatment for high-risk non-muscle-invasive bladder cancer is not a cancer drug. It
is BCG — Bacillus Calmette-Guérin, a live attenuated strain of Mycobacterium bovis
developed a century ago as a tuberculosis vaccine.

It is instilled directly into the bladder through a catheter and held there for two hours. The
bacteria provoke a violent local immune reaction against themselves, and the immune cells recruited
into the bladder wall attack the tumor as collateral. Nobody designed it to work this way. Alvaro
Morales and colleagues published the first intravesical series in the Journal of Urology in
1976, in nine patients with recurrent superficial tumors, and half a century later a
repurposed tuberculosis vaccine is still the first-line standard.

It works, and then it stops working. Reviews put the proportion of patients who eventually fail
BCG at 40 to 50%, and giving more BCG after failure produces disease-free survival of only
60 to 70% at twelve months.

The FDA has a formal definition of failure, which matters because it determines who can be
enrolled in a trial and who a drug can be sold to. In the agency’s August 2024 draft guidance,
BCG-unresponsive means persistent or recurrent CIS within twelve months of adequate BCG, or
recurrent high-grade Ta/T1 within six months, or high-grade T1 disease at the first evaluation after
induction. “Adequate BCG” is itself defined by dose count: at least five of six induction doses plus
at least two of three maintenance doses.

04The shortage nobody has fixed

There is a complication in the middle of all this that has nothing to do with biology.

The world does not have enough BCG. Merck’s TICE strain has been, since 2012, effectively
the only supply in many countries, after Sanofi Pasteur and another manufacturer hit production
problems and Sanofi never came back. Before that, Merck held roughly 28% of the US market.
The company has said its plant runs at a maximum of roughly 600,000 to 870,000 vials a year,
and that since January 2019 it has allocated the product proportionally between countries
based on historical ordering. That is not a market. That is rationing.

The FDA acknowledges it in writing. Its guidance for this indication contains a section titled
BCG Supply Issues, which permits enrolling patients who received less BCG than the definition
requires, while warning that doing so “creates uncertainty” in interpreting endpoints like durable
complete response in a single-arm trial.

The clinical effect is measurable, and the evidence is not all in one direction. A Spanish cohort
of 158 high-risk patients found that the reduced-dose schedules forced by the shortage shortened
median time to recurrence from 382 to 285 days, a hazard ratio of 1.87. A larger MD Anderson
series of 563 patients found that a third of a dose did not worsen recurrence, progression or
cancer-specific survival in patients who still received adequate BCG overall. A SEER-Medicare
analysis of 24,410 patients found no change after the 2012 shortage and a measurable drop in the
probability of receiving BCG after the 2014 one.

Read together, the picture is of a standard of care that is rationed, delivered inconsistently,
and impossible to hold constant across a clinical trial. The response rates that follow were measured against that background.

05Removing the bladder, and why patients refuse

When BCG stops working, the guideline answer has for decades been surgical: radical
cystectomy. The bladder is removed, along with nearby lymph nodes and usually the prostate or
parts of the reproductive tract, and the surgeon builds a new route for urine to leave the body,
either into an external bag or into a reconstructed internal pouch.

It is effective, and it is one of the heaviest operations in medicine. A contemporary series of
753 patients reported 2.1% mortality at 30 days and 6.9% at 90 days, with more than a quarter
of patients readmitted to hospital within 90 days. A separate series of 535 patients found
complications in 56.4% within 90 days, high-grade complications in 18.7%, and mortality of
3.9%.

Now put that against the median age at diagnosis of 73. In a German series stratified by age,
90-day mortality after cystectomy was 8.7% in patients aged 75 to 84 and 14.3% in those 85 and
over, with serious complications in a quarter to a third of them. A European expert panel
convened in 2026 concluded that many patients are simply not operable, that many others refuse, and
that the field does not even have an agreed objective definition of who counts as unfit for
surgery.

The FDA states the resulting situation plainly in its guidance: radical cystectomy is the standard
of care, and “many of these patients prefer to avoid cystectomy despite the potential risk of
progression.”

That sentence is the entire commercial rationale for the drugs that follow. They exist to give a
patient in their seventies, whose bladder cancer has survived the only treatment that reliably works,
something to try before losing the organ.

06The five approved options

Five products carry an FDA approval in this setting. They arrived in a rush, and the ordering
matters.

Valstar (valrubicin), approved in September 1998, is the fossil in the group. It is
a chemotherapy instilled into the bladder, approved for BCG-refractory CIS in patients for whom
immediate cystectomy would carry unacceptable risk. Its complete response rate was 18%, and
the label instructs physicians to tell the patient that the drug produces a complete response in
only about one in five. It also records something the newer labels do not: among the 90 patients
studied, 11% developed metastatic or deeply invasive disease, with a median delay to
cystectomy of 17.5 months.

Keytruda (pembrolizumab), from Merck, was approved for this indication on 8 January
2020 — a PD-1 checkpoint inhibitor given intravenously rather than into the bladder. In
KEYNOTE-057, among 96 patients with BCG-unresponsive CIS, the complete response rate was 41%
with a median duration of 16.2 months.

Adstiladrin (nadofaragene firadenovec), from Ferring, approved 16 December 2022, was
the first adenoviral gene therapy in this disease. It delivers a gene for interferon alfa-2b into
the bladder lining, turning the patient’s own cells into a local interferon factory for a few
months. In 98 evaluable patients with CIS: complete response 51%, median duration 9.7
months.

Anktiva (nogapendekin alfa inbakicept), from ImmunityBio, approved 22 April 2024,
is given with BCG rather than instead of it. In 77 patients: complete response 62%,
with 58% of responders past twelve months and 40% past twenty-four.

Inlexzo (gemcitabine intravesical system), from Johnson & Johnson, approved 9
September 2025, is not a drug so much as a device: a small pretzel-shaped silicone system placed
in the bladder that releases gemcitabine continuously for three weeks at a time. In 83 patients the
complete response rate was 82%, the highest number in the field.

One more is close. Cretostimogene grenadenorepvec, from CG Oncology, is an oncolytic virus
that reported a complete response rate of 75% in 110 evaluable patients in the BOND-003
trial, published in The Lancet Oncology in August 2026. It is not approved: the company
has been submitting its licence application in stages and expects to complete it by the end of
2026.

07Why those numbers are not a ranking

Line those numbers up — 18%, 41%, 51%, 62%, 75%, 82% — and they look like a leaderboard with
ImmunityBio in the middle and Johnson & Johnson in front. That reading does not hold, and the reasons matter.

None of these trials had a control arm. Every modern approval in this indication rests on
a single-arm study. There is no randomized comparison between any two of these drugs, and none
against BCG alone or against surgery.

The populations differ. Different proportions of patients with papillary tumors alongside
their CIS, different amounts of prior BCG in an era of rationing, different ages.

The measurement differs. Complete response in this disease is established by cystoscopy,
biopsy and urine cytology, and the guidance notes that results shift depending on whether a trial
used enhanced cystoscopy, whether it took mapping biopsies or targeted ones, and whether a focal
patch of CIS was removed by the screening procedure itself before treatment even started.

The trap has a documented example. In the SWOG S1605 trial, atezolizumab produced a
complete response in 20 of 74 patients with CIS, 27%, and the investigators concluded it had
failed to reach the prespecified threshold — while writing that its efficacy was “similar to results
from similar trials with other agents.” Grade 3 to 5 treatment-related events occurred in 16% of
patients, including three treatment-related deaths.

That 27% was measured at six months. Keytruda’s 41% was measured at three. They are
not the same quantity, and the difference is not a detail: a systematic review of 42 trials and
2,254 patients published in European Urology in 2020 found that across bladder-sparing
therapies the median complete response in CIS-containing disease was 26% at six months, 17% at
twelve, and 8% at twenty-four. Response rates in this disease decay steeply with time, so the
month at which a number is taken changes the number.

Read against that benchmark, atezolizumab’s 27% at six months sits exactly on the historical
median — which is what its investigators were saying when they called it similar to other agents, and also why they called it a failure.

Two further findings from the literature belong here, and neither comes from a company release.

The endpoint everything rests on is a weak proxy for survival. A meta-analysis of ten
trials published in Cancer in 2026 tested how well complete response rate predicts overall
survival in high-risk non-muscle-invasive disease. Event-free survival correlated strongly
(R² 0.72 at trial level). Complete response rate did not: R² of 0.36 at three months and
0.30 at six. The number on which every approval in this field rests explains roughly a third of
the variation in how long patients live.

An independent cost-effectiveness analysis ranks the field differently from the response
rates. The International Bladder Cancer Group modelled the approved options in European
Urology Oncology in 2024. Against radical cystectomy at a $100,000 per quality-adjusted life
year threshold, nadofaragene firadenovec was cost effective at an incremental ratio of $10,014
per QALY, while nogapendekin alfa inbakicept — Anktiva — came in at $44,602 per QALY.
Pembrolizumab dominated both, being simultaneously less costly and more effective in the model. That
analysis predates the approval of the gemcitabine system, and cost-effectiveness models are
sensitive to their price assumptions, but it is the only independent ranking of these products that
exists.

ImmunityBio has produced its own comparison. In an abstract presented at ASCO in 2026, the
company reported a matched-adjusted indirect comparison of Anktiva plus BCG against pembrolizumab,
finding a complete response at twelve months of 47.3% versus 18.8%, an odds ratio of 3.88.
The authors state the limitation themselves: it is an unanchored indirect comparison, and findings
“should be interpreted cautiously.”

And one drug did not make it at all: Vicineum, from Sesen Bio, received a Complete Response
Letter in August 2021 over both clinical-statistical questions and manufacturing findings, and was
never approved.

08Anktiva, and what its evidence consists of

ImmunityBio is a Californian company with an approved product, real revenue, and a founder who
owns most of it. Its drug, Anktiva, is an IL-15 receptor agonist: a complex of a modified
interleukin-15 molecule bound to a receptor fragment, engineered so that it activates natural killer
cells and CD8 T cells while leaving regulatory T cells — the ones that suppress immune responses —
alone.

The practical design is what makes it unusual. It is not a replacement for BCG, it is an
amplifier: 400 micrograms mixed into the BCG suspension, instilled together, held in the bladder for
two hours. Systemic exposure was below the limit of quantification in every patient tested at the
approved dose, meaning essentially none of it leaves the bladder. Treatment can run to 37 months.

The approved indication reads, exactly: Anktiva in combination with BCG “is indicated for the
treatment of adult patients with BCG-unresponsive nonmuscle invasive bladder cancer (NMIBC) with
carcinoma in situ (CIS) with or without papillary tumors.” CIS is required.

The registrational data come from QUILT-3.032, a single-arm study. The label reports 77 patients:
complete response 62%, responses ongoing past 47 months in some patients. An updated cohort of
100 patients, used for the European filing, reported a complete response rate of 71% with a
median duration of 26.6 months, and 84% of responders still free of cystectomy at three
years. The label also records that 10% of patients progressed to muscle-invasive disease.

Approvals have accumulated outside the United States: the UK in July 2025, Saudi Arabia in
January 2026, the European Union on 16 February 2026 under a conditional authorisation, and the United Arab Emirates in July 2026. The company describes that last one as the only authorisation anywhere that explicitly covers papillary disease without CIS; the underlying label text has not been independently verified here.

There is a second, separate line of evidence the company promotes heavily: the idea that Anktiva
reverses lymphopenia, the depletion of lymphocytes that follows chemotherapy and predicts poor
outcomes. Pooled data from 151 lung cancer patients showed longer survival in those whose lymphocyte
counts recovered. That is an association within a pooled non-randomized dataset, not a controlled
result, and the phase 3 trial behind part of it, QUILT-2.023, is registered as terminated with
102 patients enrolled.

09August 2024: the FDA changes its mind

Anktiva was approved on 22 April 2024. In August 2024, four months later, the FDA published
a revised draft guidance for this exact indication, and changed its position.

The 2018 guidance had blessed single-arm trials, on the reasoning that no adequate therapy
existed to serve as a comparator. The 2024 revision states that approved therapies now exist, that
therefore “an RCT is preferable”, and it sets out four specific problems with the single-arm
approach:

Without a control arm, adverse events caused by the drug cannot be separated from events caused
by the disease. In combination therapies — and Anktiva is given with BCG — the contribution of the
individual component cannot be assessed, so an ineffective component adds toxicity without
benefit. Time-to-event endpoints, the guidance says flatly, “are uninterpretable in a single-arm trial”, which removes exactly the outcomes patients care about most, such as time to cystectomy
or to progression. And variability in how trials are conducted changes the measured response rate
itself.

The agency adds a fifth problem specific to this disease: because of the BCG shortage, patients
enrolled in these trials received inconsistent and sometimes inadequate amounts of the standard
therapy, which makes a durable response rate measured in a single arm harder to interpret.

There is an irony in the sequence that the company’s critics and its supporters can both claim. The 2018 guidance permitted single-arm trials on the express reasoning that no effective therapies existed to serve as a comparator. The 2024 revision states that standard-of-care treatments now do exist, which is what makes a randomised trial possible. Anktiva is one of those treatments. ImmunityBio’s own approval helped create the condition under which the agency raised the bar.

Nothing in that document invalidates an approval already granted, and a draft guidance is explicitly not binding on the FDA or on anyone else. But it describes the terms on which the next application will be judged, and ImmunityBio’s application for papillary-only disease is the next one.

10The refusal, and the road to 6 January 2027

What happened next is easier to follow in order, because the company’s account and the agency’s conduct do not sit comfortably together.

In March 2025 ImmunityBio submitted a supplemental application for papillary-only disease,
based on Cohort B of QUILT-3.032: 80 patients, disease-free rate at twelve months of 58.2%.
The company states that it filed after a January 2025 in-person meeting at which FDA leadership
“unanimously confirmed” it should submit on the strength of the single-arm data.

On 2 May 2025 the FDA issued a refuse to file letter. This is not a rejection of a
drug; it is a refusal to review the application at all, on the grounds that it is not complete
enough to examine. The company’s own press release described it as an “inconsistent refusal to file.”
At a Type A meeting in June, per the company, the agency explained the refusal rested on requiring a
randomized controlled trial against chemotherapy.

A Type B meeting followed in December 2025. In January 2026 the company announced
that the FDA had asked for additional information and, notably, that this “does not contemplate the
initiation or design of a new clinical trial.” The information went in during February. In March the
agency asked for updated efficacy data, and the company resubmitted.

On 19 May 2026 the FDA accepted the application for review and set a PDUFA target action
date of 6 January 2027. The acceptance letter, quoted by the company, describes what the review
will actually turn on: ImmunityBio has submitted “a literature-based rationale proposing that
papillary NMIBC has an overlapping clinical and non-clinical profile with CIS that may allow for
extrapolation of results from patients with CIS,” and the scientific data behind that overlap “will
be the focus of the review.” The same announcement records that the agency reiterated “their
concerns relating to single-arm trials in papillary disease alone.”

So the question in front of the FDA on 6 January 2027 is not really whether Anktiva works in
papillary disease. It is whether a body of evidence built in CIS patients can be stretched to cover
patients who do not have CIS — decided by an agency that spent the previous two years writing down
why it no longer likes the study design that evidence came from.

Two further details belong to the record. The NCCN guidelines added Anktiva plus BCG
for papillary-only disease in March 2026 as a Category 2A recommendation, which the company
itself describes as covering “a use that is not included in the current FDA approved indication” —
in other words, off-label. And neither the January 2026 announcement nor the May 2026 acceptance was
filed as an 8-K; both exist only as press releases, with SEC confirmation arriving later inside the
quarterly filings.

11Real revenue, a paper loss, and a founder on both sides

ImmunityBio has something most clinical-stage biotech companies do not: a product on the market that people are buying.

Net product revenue was $50.7 million in the second quarter of 2026, up 92% year on
year and 15% sequentially — the eighth consecutive quarter of sequential growth. First-half revenue
was $94.8 million against $42.9 million a year earlier. Full-year 2025 revenue was $113.0
million. Cash and marketable securities stood at $357.4 million on 30 June 2026.

Then there is the headline loss, which misleads unless it is broken apart. Net loss
attributable to shareholders was $863.2 million in the first half of 2026, against $222.2
million a year earlier. That is not an operating collapse. Loss from operations was $131.5
million and actually improved year on year. The difference sits almost entirely in two
non-cash items: a $332.6 million revaluation of a convertible note and a $339.6
million revaluation of warrant liabilities. Both grew because the share price rose.

The balance sheet is where the story gets structural. Total liabilities of $1.674 billion sit against total assets of $628.5 million, and the deficit attributable to ImmunityBio shareholders is negative $1.047 billion. The
10-Q states that “substantial doubt exists regarding our ability to continue as a going concern
without additional funding or financial support,” and describes that doubt as alleviated partly by
the founder’s “intent and ability to support our operations with additional funds.”

That founder is Patrick Soon-Shiong, and his position is unusual on both sides of the
balance sheet. His most recent Schedule 13D/A reports beneficial ownership of 745,518,285 shares, or 64.8%, as of 31 July 2026 — a figure that rests on a diluted share count rather than on shares outstanding, and which the company’s own proxy puts at about 62.5% on a different basis and date. Either way ImmunityBio is a “controlled company” under Nasdaq rules. He is also, through Nant Capital, the holder of a
convertible note with $480 million of principal outstanding, carrying interest at SOFR plus
8%, maturing 31 December 2027, convertible at $5.4270 a share. In January 2026 the note was amended to let him convert at any time; in March he converted $25 million into 4.6 million shares. That $480 million is the principal still outstanding; because the note is carried at fair value, it appears on the balance sheet at $774.4 million, and the gap between the two figures is what produced the revaluation charge.

A separate obligation sits ahead of him in the queue. Under a revenue interest agreement with
Oberland Capital, the company has received $375 million against tiered royalties of
5.625% to 12.5% of worldwide net sales, secured by a pledge over substantially all assets.
The carrying value of that liability was $415.1 million at 30 June 2026. Financing has also come
from an at-the-market programme that produced $147.2 million in the first half, leaving $349.2
million still available, and from two registered direct offerings in 2025 that raised roughly
$155 million gross together.

Shares outstanding reached 1,053,221,645 at 30 June 2026, up from 1,011,800,008 six months
earlier. At around $8.50 a share during the session of 27 August 2026, that puts market capitalisation near $9.0 billion — a figure that moves with the price and will have changed by the time you read it.

12The rest of the pipeline, and the bigger market behind it

Anktiva is not the only programme in the company, though it is the only one generating revenue.

The BCG-naive trial matters more than it looks. QUILT-2.005 tests Anktiva plus BCG in
patients receiving BCG for the first time, rather than after it has failed. It is fully enrolled at
369 patients, with a pivotal readout guided to the second half of 2026 and a licence application
targeted between late 2026 and early 2027. That population is far larger than the BCG-unresponsive
one, and AstraZeneca has already moved into it: Imfinzi plus BCG was approved for BCG-naive
high-risk disease on 28 May 2026, on the strength of POTOMAC — a randomized trial in 1,018
patients. Whatever ImmunityBio brings to that market will be judged against a randomized
comparison.

Supply is being addressed directly. In May 2026 the company signed an exclusive
development and supply agreement with Japan BCG Laboratory for the Tokyo-172 strain, with no
payments due before FDA approval, and it has an expanded access programme running with recombinant
BCG from the Serum Institute of India, under which it says more than 570 patients have been treated
at 58 centres. A company whose product must be mixed with BCG has an obvious interest in there being
BCG.

Lung cancer is the second pillar, and the least well documented. ResQ201A is a phase 3 trial
in non-small-cell lung cancer resistant to checkpoint inhibitors, with overall survival as the
primary endpoint. The registry lists a target of 507 patients; the annual report describes 462
randomized 2:1. Those are not the same number, and the company has not reconciled them.

Several programmes have been shut down over the past year, which is normal in drug development: a
phase 3 in papillary NMIBC, a phase 3 in first-line lung cancer, and a CD19 cell therapy trial in
leukaemia all appear on the registry as withdrawn or terminated.

13What to watch, and what each signal means

The next twelve months come down to a small number of observable things.

6 January 2027. The PDUFA date for papillary-only disease. Approval would make Anktiva the
first drug in the world with a US label covering that population, in a field where four competitors
are locked into CIS. Refusal leaves the papillary use where it is now: in the NCCN guidelines,
off-label, and unbillable as an approved indication.

What the decision actually rests on. Not efficacy in the ordinary sense, but whether
evidence generated in CIS patients can be extrapolated to patients without CIS. The FDA said so
in writing, and said in the same breath that it retains concerns about single-arm data in this
population.

The BCG-naive readout in the second half of 2026. A larger market, a comparator already
approved on randomized evidence, and a company whose entire regulatory history so far runs on
single-arm trials.

Revenue trajectory. Eight consecutive quarters of sequential growth is a real
achievement in a product that has to be mixed with a rationed component and instilled by a
urologist. The number to watch is whether growth continues once the initial wave of BCG-unresponsive
patients has been treated, since each patient’s course is finite.

The founder’s note, and the clock on it. $480 million maturing on 31 December 2027,
convertible at $5.4270, held by a man who already owns about 65% of the shares and can now convert at
any time. Conversion removes debt and adds dilution; repayment removes cash. Either path shows up in
the share count or the balance sheet, and the 10-Q’s going-concern language ties the company’s
solvency explicitly to his willingness to keep funding it.

The undrawn ATM. $349.2 million of capacity remained at 30 June 2026. The share count in each quarterly filing is where any use of it becomes visible.

The through-line, if there is one, is that ImmunityBio built a real commercial product by
getting through a regulatory door in April 2024 that the FDA began closing four months later. The
company is now standing at the same door with a second application, arguing that data from one group
of patients should count for another. That argument gets decided on 6 January 2027, and everything
else — the revenue, the valuation, the founder’s balance sheet — is downstream of it.