Passed Phase 3, Blocked Three Times at Approval: Dissecting Rivoceranib's FDA Gate

A cancer that starts in the liver grows by recruiting new blood vessels. Rivoceranib is a pill that switches off the machinery building those vessels — an oral targeted therapy (TKI) that blocks the vascular endothelial growth factor receptor (VEGFR2) to cut off a tumor's blood supply. It is the same molecule known in China as apatinib, and it was originally developed there. To this you add camrelizumab, an injected drug that releases the immune cell's brake — an immunotherapy that undoes the lock (PD-1) keeping immune cells from attacking the cancer. Used together, the combination beat the control arm in first-line treatment of unresectable hepatocellular carcinoma (HCC), and it was already approved in China back in January 2023.

In the United States, though, it was blocked three times. May 2024, March 2025, and July 2026 — the FDA sent this combination's new drug application back three times.

"A drug that succeeded in Phase 3 was blocked three times in the US" — the sentence pushes you straight toward a shortfall in efficacy. If the trial passed but the review didn't, the assumption goes, the data must have been weak or the FDA must have flagged safety. But the disclosed reasons point elsewhere. Drug approval turns on more than efficacy alone, and the point where this drug got stuck was a factory outside the United States. What is more, the fact that the only voice reporting this is the party that wants the approval is itself the trickiest variable in the story. This piece takes that gate apart — what is actually deciding rivoceranib's entry into the US market, and how, then, should we read the odds on a fourth attempt.

Approval Is Not a Single Gate

For a new drug to enter the US market, it has to clear three gates at once. Does the drug actually work (efficacy), can it be made at the same quality every time (manufacturing), and do those data and facilities hold up under inspection (on-site inspection). If even one of the three is left open, no approval comes. In logic-circuit terms, it is an AND gate — the output is true only when every input is true.

The rivoceranib combination cleared the first gate. Its Phase 3 results stand as independent evidence vetted by an international journal (the figures come in the next section). The problem is the other two gates.

When the FDA says "we can't approve it as is" and asks for fixes, it sends a complete response letter (CRL). It is neither approval nor rejection — a hold tag. The company can fix what was flagged and resubmit, and the review then resumes. The rivoceranib combination has received this tag three times.

All three disclosed reasons landed on the manufacturing and inspection side.

First CRL (May 2024). The first letter cited unfinished manufacturing and quality work (CMC, chemistry, manufacturing and controls) and on-site inspection (BIMO, Bioresearch Monitoring). CMC is a review of the factory, not the efficacy lab — it asks whether the drug can be stamped out at identical quality every time. BIMO is a different kind of thing: at its core, it is FDA auditors going to the clinical trial sites in person to verify the integrity of the trial data (the pre-approval inspection that checks a factory's manufacturing compliance is a separate procedure). A CMC problem at a Hengrui manufacturing facility coincided with COVID-19 travel restrictions that had left the data inspection of some clinical sites unfinished.

The company said that, after a re-inspection, it had passed this on-site inspection. It appeared to have cleared both the first gate and the inspection gate.

Second CRL (March 2025). But the next letter raised a new finding. This time it was an unresolved manufacturing (cGMP) lapse at a plant owned by China's Hengrui — the target was the manufacturing process for the partner drug, camrelizumab. The earlier inspection had passed, yet a different manufacturing point got flagged. The bottleneck hadn't disappeared so much as moved.

Third CRL (July 2026). The third letter's reason was in the same family, but the point that got flagged shifted again. This time it was not camrelizumab but the factory making rivoceranib's own active ingredient. A Hengrui facility in China, listed on the rivoceranib new drug application, received a Form 483 at a routine FDA cGMP inspection in April 2026. A Form 483 is the "list of things to fix" that an auditor leaves after an inspection. The FDA required the company to resolve this deficiency and demonstrate compliance before approval, and gave notice that a further inspection might follow.

That is as far as the FDA's own wording goes — it said only "a manufacturing site listed on the rivoceranib new drug application," without specifying which product line at that facility was the problem. Korean reporting fills that gap. According to that reporting, the facility both makes rivoceranib's active pharmaceutical ingredient (the apatinib drug substance) and doubles as a shared plant that also stamps out the active ingredient for another generic drug sold in the US, and the inspection that triggered this 483 was a routine check of that generic ingredient line, which has nothing to do with rivoceranib — a check the company says it did not know about in advance. In other words, rivoceranib's own approval got tripped up by a routine inspection of an unrelated drug under the same roof.

All three times, by the company's own disclosure, the reason was not efficacy. But the point that got flagged differed each time — the first and second were camrelizumab's manufacturing process, the third was rivoceranib's own ingredient facility. There was one thing in common. All three sat under the same roof: China's Hengrui.

Here is where the story gets strangely structured. The full text of an FDA CRL is not made public. So the only voices who can say "efficacy was not the issue" are the developer, Elevar, and HLB, and their CEO — all interested parties. When the stock drops to its daily floor every time a CRL lands, framing the reason as a manufacturing problem works in their favor. In other words, the testimony backing this piece's core diagnosis — that efficacy is irrelevant — comes solely from the mouths of the parties who stand to gain most from that diagnosis. So this piece does not write "the FDA did not take issue with efficacy" as a factual proposition; it writes it only as an attributed statement — that this is so by the company's own disclosure. This asymmetry of sourcing comes back, more heavily, when we weigh the odds at the end.

To sum up the three CRLs: the efficacy gate opened (passed). What got flagged was the manufacturing and inspection gate every time, and the flagged point wandered from camrelizumab toward rivoceranib. But there was one point it could not escape however far it wandered — the single-factory bottleneck that is China's Hengrui. Rivoceranib's entire entry into the US is tied to this one exogenous bottleneck, and the bottleneck moves between the two drugs. This is the pressure point of the chain.

Here the asymmetry between the gates shows itself. Efficacy is a constant whose value is already fixed — Phase 3 is over and the figures are set in the paper. Manufacturing and inspection, by contrast, are a variable that has to clear one point, Hengrui, again and again. It is a point the company cannot control, and it is exogenous enough that even a routine inspection of a generic unrelated to this drug can block this drug's approval. However good the efficacy, if this variable is left open, no approval is possible.

Was Efficacy Really Not the Gate?

For the diagnosis that "manufacturing is the gate" to hold, we first have to nail down that the efficacy gate was actually passed. If the CRL reasons had efficacy or clinical findings mixed in, the thesis that "manufacturing is the bottleneck" collapses.

Nothing in the public record confirms such a finding. One oncology trade outlet, citing the company's announcement, reported that neither the first nor the second CRL flagged any concern about the combination's clinical efficacy or safety. The third CRL was the same — according to the company, the FDA did not raise concerns about efficacy or safety or request an additional clinical trial, and the CEO stated that "no deficiency was identified related to clinical efficacy and safety data." The company also stated that the FDA had earlier acknowledged "substantial clinical data" for this application.

But this is as far as the company's announcement reaches. What these statements tell us is not "the efficacy is flawless" but only "efficacy is absent from the disclosed reasons."

Evidence for the efficacy gate itself sits independently, outside the company's announcement. The clinical basis is the CARES-310 Phase 3 trial. An international, open-label trial that randomized 543 patients with untreated, unresectable hepatocellular carcinoma across 95 institutions in 13 countries, it pitted the rivoceranib-camrelizumab combination one-to-one against sorafenib. The results were published in the Lancet family of international journals.

Two terms before reading the table below. The hazard ratio (HR) is a relative ratio comparing the risk of an event between two arms; below 1 means the risk fell in the treatment arm. And this trial was a "dual primary endpoint" design that pre-specified both overall survival (OS) and progression-free survival (PFS) as primary endpoints — the condition being that both, not one or the other, must succeed.

Source: CARES-310 Phase 3 final analysis (Lancet Oncol 2025 · Lancet 2023) · as of 2025. A hazard ratio of 0.64 means a 36% reduction in the risk of death, one-sided p<0.0001.

On overall survival, the combination's 23.8 months led sorafenib's 15.2 months by 8.6 months. A hazard ratio of 0.64 means the risk of death was 36% lower, and it was statistically significant. On progression-free survival, too, the combination led (hazard ratio 0.54). Both dual primary endpoints were met. These figures sit in a peer-reviewed paper, not in an interested party's press release. This is the bedrock of the diagnosis that the efficacy gate was passed.

One caveat to attach. CARES-310's control arm, sorafenib, was the first-line standard when the trial was designed (around 2019), but in the meantime the US first-line standard for hepatocellular carcinoma moved to immune combinations like atezolizumab plus bevacizumab and durvalumab plus tremelimumab. The nivolumab-plus-ipilimumab combination was also approved for first-line treatment in April 2025. So "passed the efficacy gate" means, precisely, that the endpoints were met versus sorafenib, not that superiority over the current standard was demonstrated. Even if the manufacturing bottleneck clears, where this drug would stand in today's first-line market remains a separate clinical question.

Why the Bottleneck Sits Beyond Control

One could object: CMC and inspection are merely procedural defects, so as long as the efficacy is solid, a re-inspection will eventually get it through, and calling it a "bottleneck" is an overstatement. But it is a bottleneck precisely because it is procedural.

Efficacy is already fixed. Phase 3 is over and the figures are set in the paper. Inspection, by contrast, hinges on variables like facility access, inspector scheduling, and the re-inspection queue — conditions the company cannot control. It doesn't get resolved by submitting more data. It gets resolved only when an auditor goes back to the factory and confirms the corrective items are closed. That "when it gets resolved" depends on scheduling and access rather than data — that is what the bottleneck really is. This diagnosis of a procedural problem, though, has an expiry date. When the same procedure repeats three times, you start to suspect something beyond procedure — a point we return to at the end.

The three CRLs we saw show this property exactly. Even after the first inspection passed, the second brought a new finding at camrelizumab's manufacturing point, and the third snagged a Form 483 at a routine inspection of rivoceranib's ingredient facility. The point that gets flagged keeps moving. It isn't like efficacy, settled once and done.

And this bottleneck is tied to geopolitics. Both the place that makes camrelizumab and the place that makes rivoceranib's ingredient are Hengrui facilities in China. Hengrui is camrelizumab's original developer and the party responsible for its manufacturing and licensing application, and rivoceranib's ingredient is likewise made at a Hengrui facility. The US-based Elevar, which handles rivoceranib, and its parent company HLB do not directly control these factories.

A common objection here goes: if the troubled facility belongs to the partner, isn't this not rivoceranib's own problem but merely "partner risk"? But the third CRL erases that very distinction. What got flagged in the third was not the partner drug camrelizumab but the facility making rivoceranib's own ingredient. So this is neither a "partner problem" nor an "own problem." The fact that both drugs are made at a single place — China's Hengrui — that is, the dependence on Chinese manufacturing itself, is the gate. The structure being dissected is that rivoceranib's gate is tied not to its own efficacy but to this single-factory bottleneck. Laying out where each party stands makes it clear why the bottleneck gathers in one place.

Source: Elevar and Hengrui disclosures (globenewswire · elevartx) · Korean reporting (edaily) · as of 2026-07.

The stock moves in South Korea, the new drug application gets stuck in the US, and the inspection bottleneck binds both drugs to a single factory in China. In this structure spanning three countries, the pressure point that decides approval is an exogenous variable the Korean listed company cannot control — exogenous enough that even a routine inspection of a generic unrelated to this drug can become an obstacle.

What Did the Market Misread?

Now back to the opening mystery. The question "it succeeded in Phase 3, so why won't approval come?" was, in fact, a question that saw only efficacy as the gate. Unless the other two gates come into view, that question doesn't resolve.

Where this gap in view shows up most sharply is the stock. HLB's stock reacted to all three CRLs with a near-limit-down drop each time.

Source: Korea Exchange · KED Global · Seoul Economic Daily · based on each date's closing price.

At the first CRL, the affiliates too sank to their limit-down floor in unison, and more than KRW 5 trillion of the group's market capitalization evaporated in a single day. At the second, some KRW 3 trillion vanished in a day across the group's 10 companies. At the third, five of the group's companies headed straight to limit down right after the open.

Two things to concede first. First, the mere number of roughly -30% all three times cannot prove that "the market failed to distinguish the reasons." The Korean market has a price limit capping a day's drop at ±30%, so when strong bad news like a CRL lands, it hits the floor first whatever the reason. The three drops looking identical may be a ceiling the system pressed down rather than the market's judgment being identical. Second, the plunge itself may be rational. As approval is postponed three times, the US revenue opportunity recedes accordingly, and resubmission costs and capital-raising terms worsen. This is not a misread of the narrative but a legitimate repricing of the fundamentals.

So this piece will not argue that "the stock fell the same three times, therefore the market misread it." The size of the drop does not prove a misread.

Still, one observation survives, narrowly, on top of that. It is not the stock itself but the discourse around it — every time a CRL lands, the frame "a Phase 3-successful drug was rejected again" repeats, and that frame erases what it was blocked over. The fact that the problem was manufacturing and inspection, not efficacy — and manufacturing at a different point each time — mostly drops out of the narrative. To an eye that reads efficacy as a binary, "Phase 3 success = approval imminent," the gate that is the manufacturing bottleneck is invisible from the start. What splits here is not the stock's right or wrong but the reality science points to (the manufacturing and inspection bottleneck) versus the picture the popular narrative draws (efficacy either passes or fails). HLB's stock is only material that lights up this gap, not the protagonist of this piece.

So, What Are Rivoceranib's Approval Odds?

The dissection has pinpointed the pressure point. Whether the rivoceranib combination is re-approved turns not on the efficacy narrative but on whether the inspection bottleneck at a single factory — China's Hengrui — clears. So how should we read the odds on a fourth attempt?

First, let's set out what would show the diagnosis right and what would show it wrong. Instead of predicting a direction, here are the indicators.

• Leading signals (toward the bottleneck clearing): completion of a re-inspection at Hengrui's Chinese facility · confirmation that the Form 483 findings are closed · passing an additional pre-approval inspection.
• The quality of the next resubmission: how cleanly the same applicant closes the facility issues before filing again this time. Given the precedent of declaring "full resolution" only to be snagged again five months later, the completeness of the resubmission is itself a window into the company's quality system.
• The geopolitics of US-China manufacturing inspection: the FDA's on-site inspections in China hinge on bilateral relations, the inspection backlog, and the decoupling trend. Macro variables unrelated to this drug could govern the timing of the re-inspection.
• Disconfirming signals (toward the diagnosis weakening): a fourth CRL even after resubmission, or the review drifting for a long time without a PDUFA. If that happens, the "simple manufacturing bottleneck" diagnosis wobbles, and the problem shifts to a deeper layer such as facility trust or data integrity.

This piece sets the odds low. This is not a confident prophecy but a low-confidence judgment. And the basis for "low" is not uncertainty itself but the track record. If the inspection has been blocked three times running, and the declaration that "the previously raised concerns are fully resolved" was overturned five months later, that is in itself evidence of a persistent flaw in this company's and facility's quality system. It is not simply that "we don't know the outcome" — the execution record so far points to a low probability.

On top of that, two counterexamples press the odds down further — the passage where this piece hits itself hardest.

First, the only source for the claim that efficacy is not the gate is an interested party. This is exactly the problem this piece flagged at the outset. The full CRL text is not public, and the only voice for the statement "efficacy was not the issue" is a single company with a clear motive to defend its stock. The possibility that the company framed an efficacy-related issue as a manufacturing issue cannot be independently verified from public information alone. This counterexample cannot be knocked down — it is the kind you have to concede and move past. Yet one thing limits its force. The efficacy Phase 3 itself stands as independent evidence published in the Lancet family of journals, not in a company announcement. So this piece claims only that "the disclosed reason is manufacturing and efficacy passed on independent evidence," and does not claim that "efficacy is flawless and irrelevant." This limit of sourcing is the first basis for setting the odds low.

Second, the sheer repetition of three times. At its February 2026 resubmission, the company stated it had "fully resolved the previously raised concerns." Yet five months later — and earlier than the review deadline (the PDUFA date, the day the FDA has promised to deliver its decision) of July 23 — the third CRL arrived. The claim of full resolution was overturned within five months. Three in a row, and a CRL earlier than the scheduled date at that, is hard to explain by "a simple procedural defect" alone. The repetition and the early notice point to a more fundamental problem, such as facility trust or data integrity.

So I lower the diagnosis a notch. From "manufacturing is the bottleneck" to: "the disclosed reason is manufacturing and inspection, but this very repetition — three in a row, an early notice, and a full-resolution declaration reversed — is spilling beyond simple procedure into a trust problem in the quality system." It also remains a limitation that, from the information now public, we cannot tell how serious the Form 483 findings are. This repetition is the second basis for setting the odds low.

One thing to add: these three years of drift do not mean a "treatment gap" for US patients. As we saw, in the meantime several approved alternatives, atezolizumab plus bevacizumab among them, have taken hold in the US first-line hepatocellular carcinoma market. It is not a situation where, because this drug can't get in, patients have nowhere to turn. What is at stake here is one company's market entry rather than patients' urgency.

To sum up: what decides rivoceranib's entry into the US is not efficacy but the manufacturing and inspection bottleneck at a single factory, China's Hengrui, and that bottleneck has moved between the two drugs and repeated three times. The source for the claim that the reason is purely manufacturing is a single interested party, and that very point is the trickiest variable in the story. This piece reads the odds on a fourth attempt as low — not out of pessimism, but because the record of three failures and a reversed declaration points that way. The indicator to watch is not efficacy data but the re-inspection result at that factory. The Phase 3 figures are already what we know; what remains to be known is what the auditor confirms at that factory.

Note: This piece does not recommend buying or selling any particular security; it is information and regulatory analysis based on publicly available materials. Investment decisions and their consequences rest with the investor.