Biotech iQ

$REPL I opened my position in REPL after the second CRL at under $2. This morning, after going through both the FDA and company briefing documents, I decided to close my position and move to the sidelines. When $URGN was in a similar spot, my read was that the quality of the data supported management's arguments. I doubled down after the AdCom despite the negative vote. It was a risk, but in that case, it worked out. With Replimune, the FDA went far beyond questioning a single-arm trial design: it stated (rightfully so IMO) that REPL's representation to the agency was not true--specifically, Replimune's claim that no biopsy or excision of a lesion remnant resulted in an improved best overall response, and its separate assertion that histology-based reclassification did not apply to this cohort. But an even bigger issue is, does the data answer the right question? Does the data demonstrate that RP1 provides benefit as a systemic therapy? I think the data show clearly that injecting an oncolytic herpes virus into a tumor shrinks that tumor. I don't think that was ever in dispute. Direct oncolysis, needle trauma, injected volume, and local inflammation have already been shown to produce a local effect. But local shrinkage is not the question. The question is whether the response reflects a systemic, disease-modifying effect. A nodule shrinking under the needle does not extend life. That is precisely what happened to Imlygic, which was narrowed after its 2015 advisory committee to local treatment, with a labeled limitation of use stating it had not been shown to improve overall survival or affect visceral metastases. (Incidentally, Imlygic and RP1 share the same founders.) Replimune's evidence for a systemic effect is the lesion-level analysis: among 108 patients with both injected and non-injected measured lesions, response based on non-injected lesions alone was 28.7% against 31.5% for injected lesions. The problem is that injection status was not assigned at baseline and left alone. RP1 was dosed every two weeks for up to eight cycles, capped at 10 mL per visit, and the protocol instructed investigators to inject the largest and then next-largest injectable tumor until the volume ran out or no injectable tumors remained. New tumors were eligible. Different tumors could be injected on different days. There was no minimum lesion size. First, a lesion that starts growing becomes a priority injection target and is thereafter coded as injected. FDA found the dataset captured only whether a lesion was ever or never injected, so a lesion injected later is classified as injected retroactively. Progressing lesions were systematically drained out of the non-injected pool over the course of treatment. Second, the volume cap and the largest-first rule mean small lesions were disproportionately skipped — and small lesions are far more likely to be scored as complete disappearance than large ones. Injection status is therefore endogenous to outcome and cannot be considered an independent variable. The FDA found that 49% of patients the company deemed responders had all of their target lesions injected, and two more had no target lesions at baseline. So over 50% of reported responders had no non-injected target lesion by which to assess systemic activity. Then there is the disease-burden gradient. At a baseline sum of diameters under 20 mm, ORR was 63.2% with a 57.9% complete response rate. Above 150 mm, ORR was 11.8% with no complete responses at all. Every CR in the trial occurred below 100 mm. If RP1 were driving systemic immunity, my view is that the CR rate should not vanish entirely as burden rises. Then there's the contribution-of-components problem that still sits on top of all of this and remains unchanged since the first CRL. There is no RP1 monotherapy arm. Every responder also received nivolumab. In any case, anything can happen--this is biotech after all. But I'm on the sidelines for this one. 🍿